0485adc · crush · git.iko.soy

git.iko.soy/crush

chore(tests): update golden files

commit0485adc4776f82c2b44a6fae30187e871ad487b8 authorKieran Klukas <kieran@dunkirk.sh> dateTue 12 May 2026, 12:04 (5 months ago) parent9b8b888

13 files changed+3102−3333

compare with main

unifiedsplit

internal/agent/testdata/TestCoderAgent/glm-5.1/bash_tool.yaml+254−248

@@ -25,351 +25,371 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summar"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"izing"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ≤"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" named"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"hello"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'."},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" do"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" not"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" print"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" its"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" timestamp"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Create"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ≤"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" using"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" without"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" timestamp"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"col"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Maybe"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"B"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ash"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" English"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" command"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Maybe"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"B"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ash"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"?"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"B"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"B"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ash"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ash"},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" command"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

177177

178− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

178+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

179179

180− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ="},"finish_reason":null}]}

180+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

181181

182− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" let's"},"finish_reason":null}]}

182+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ="},"finish_reason":null}]}

183183

184− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

184+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" let's"},"finish_reason":null}]}

185185

186− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

186+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

187187

188− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Bash"},"finish_reason":null}]}

188+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

189189

190− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

190+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Bash"},"finish_reason":null}]}

191191

192− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

193193

194− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

194+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

195195

196− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" +"},"finish_reason":null}]}

196+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")+"},"finish_reason":null}]}

197197

198− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

198+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

199199

200− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

200+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

201201

202− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

202+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

203203

204− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

205205

206− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

207207

208− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" command"},"finish_reason":null}]}

209209

210− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

211211

212− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

213213

214− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

215215

216− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"12"},"finish_reason":null}]}

217217

218− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

219219

220− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"11"},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

221221

222− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

223223

224− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

224+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"13"},"finish_reason":null}]}

225225

226− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

227227

228− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"12"},"finish_reason":null}]}

228+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

229229

230− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

230+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

231231

232− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

232+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

233233

234− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

234+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

235235

236− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"15"},"finish_reason":null}]}

237237

238− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

239239

240− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

241241

242− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

243243

244− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"20"},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"16"},"finish_reason":null}]}

245245

246− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

247247

248− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

249249

250− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

250+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

251251

252− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"21"},"finish_reason":null}]}

252+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

253253

254− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

254+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

255255

256− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

256+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"22"},"finish_reason":null}]}

257257

258− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

258+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

259259

260− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

261261

262− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

263263

264− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"25"},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"23"},"finish_reason":null}]}

265265

266− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

267267

268− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

268+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

269269

270− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

270+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

271271

272− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"26"},"finish_reason":null}]}

272+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

273273

274− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

274+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

275275

276− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

276+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

277277

278− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

278+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"31"},"finish_reason":null}]}

279279

280− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

280+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

281281

282− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

282+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

283283

284− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"31"},"finish_reason":null}]}

284+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

285285

286− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

286+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"32"},"finish_reason":null}]}

287287

288− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

288+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

289289

290− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

290+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

291291

292− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"32"},"finish_reason":null}]}

292+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

293293

294− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

294+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

295295

296− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

296+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

297297

298− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

298+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"36"},"finish_reason":null}]}

299299

300− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

300+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

301301

302− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

302+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

303303

304− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"36"},"finish_reason":null}]}

304+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

305305

306− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

306+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"37"},"finish_reason":null}]}

307307

308− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

308+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

309309

310− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

310+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

311311

312− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"37"},"finish_reason":null}]}

312+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

313313

314− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

314+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

315315

316− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

316+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

317317

318− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

318+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"42"},"finish_reason":null}]}

319319

320− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

320+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

321321

322− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

322+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

323323

324− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"44"},"finish_reason":null}]}

324+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

325325

326− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

326+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"43"},"finish_reason":null}]}

327327

328− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

328+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

329329

330− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

330+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" bash"},"finish_reason":null}]}

331331

332− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"44"},"finish_reason":null}]}

332+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

333333

334− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

334+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

335335

336− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

336+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

337337

338− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

338+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"47"},"finish_reason":null}]}

339339

340− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

340+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

341341

342− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

342+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

343343

344− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

344+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

345345

346− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

346+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"47"},"finish_reason":null}]}

347347

348− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

348+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

349349

350− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

350+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

351351

352− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"B"},"finish_reason":null}]}

352+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

353353

354− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"ash"},"finish_reason":null}]}

354+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

355355

356− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" create"},"finish_reason":null}]}

356+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

357357

358− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" test"},"finish_reason":null}]}

358+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/"},"finish_reason":null}]}

359359

360− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}]}

360+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"col"},"finish_reason":null}]}

361361

362− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" with"},"finish_reason":null}]}

362+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

363363

364− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" hello"},"finish_reason":null}]}

364+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

365365

366− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" bash"},"finish_reason":null}]}

366+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

367367

368− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" content"},"finish_reason":null}]}

368+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

369369

370− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

370+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"B"},"finish_reason":null}]}

371371

372− data: {"id":"chatcmpl-faa1080788595f33fa9510f3","object":"chat.completion.chunk","created":1778597310,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":31,"completion_tokens":180,"total_tokens":403,"prompt_tokens_details":{"cached_tokens":192}}}

372+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"ash"},"finish_reason":null}]}

373+

374+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" command"},"finish_reason":null}]}

375+

376+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

377+

378+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" create"},"finish_reason":null}]}

379+

380+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" test"},"finish_reason":null}]}

381+

382+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}]}

383+

384+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" with"},"finish_reason":null}]}

385+

386+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" hello"},"finish_reason":null}]}

387+

388+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" bash"},"finish_reason":null}]}

389+

390+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

391+

392+ data: {"id":"chatcmpl-c4e5208ce189f5ae587688b7","object":"chat.completion.chunk","created":1778601691,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":31,"completion_tokens":190,"total_tokens":413,"prompt_tokens_details":{"cached_tokens":192}}}

373393

374394 data: [DONE]

375395

@@ -378,15 +398,15 @@ interactions:

378398 - text/event-stream

379399 status: 200 OK

380400 code: 200

381− duration: 894.7825ms

401+ duration: 452.47ms

382402- id: 1

383403 request:

384404 proto: HTTP/1.1

385405 proto_major: 1

386406 proto_minor: 1

387− content_length: 33684

407+ content_length: 33771

388408 host: ""

389− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'. do not print its timestamp\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

409+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'. do not print its timestamp\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

390410 headers:

391411 Accept:

392412 - application/json

@@ -403,87 +423,55 @@ interactions:

403423 content_length: -1

404424 uncompressed: true

405425 body: |+

406− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

407−

408− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Create"},"finish_reason":null}]}

409−

410− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

411−

412− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" file with"},"finish_reason":null}]}

413−

414− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" echo"},"finish_reason":null}]}

415−

416− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

417−

418− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" suppress"},"finish_reason":null}]}

419−

420− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" timestamp"},"finish_reason":null}]}

421−

422− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" printing"},"finish_reason":null}]}

426+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

423427

424− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

428+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_1a51a6006cb9443db0b7bcb2","type":"function","index":0,"function":{"name":"bash","arguments":""}}]},"finish_reason":null}]}

425429

426− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_b10ffb78b6e844ce91c60f38","type":"function","index":0,"function":{"name":"bash","arguments":""}}]},"finish_reason":null}]}

430+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

427431

428− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

432+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

429433

430− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

434+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"command\": "}}]},"finish_reason":null}]}

431435

432− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"command\": "}}]},"finish_reason":null}]}

436+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"touch -"}}]},"finish_reason":null}]}

433437

434− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"printf"}}]},"finish_reason":null}]}

438+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"t"}}]},"finish_reason":null}]}

435439

436− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" '%"}}]},"finish_reason":null}]}

440+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" 19"}}]},"finish_reason":null}]}

437441

438− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"s' '"}}]},"finish_reason":null}]}

442+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"7001"}}]},"finish_reason":null}]}

439443

440− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello bash"}}]},"finish_reason":null}]}

444+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"010"}}]},"finish_reason":null}]}

441445

442− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"' \u003e /"}}]},"finish_reason":null}]}

446+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"000 test"}}]},"finish_reason":null}]}

443447

444− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}]}

448+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}]}

445449

446− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

450+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \u0026\u0026 echo '"}}]},"finish_reason":null}]}

447451

448− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

452+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello bash"}}]},"finish_reason":null}]}

449453

450− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

454+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"' \u003e test"}}]},"finish_reason":null}]}

451455

452− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

456+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt\""}}]},"finish_reason":null}]}

453457

454− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

458+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

455459

456− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

460+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"description\": "}}]},"finish_reason":null}]}

457461

458− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

462+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"Create test"}}]},"finish_reason":null}]}

459463

460− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

464+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt with content"}}]},"finish_reason":null}]}

461465

462− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1/bash"}}]},"finish_reason":null}]}

466+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" and zero"}}]},"finish_reason":null}]}

463467

464− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}]}

468+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" timestamp\""}}]},"finish_reason":null}]}

465469

466− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/test.txt"}}]},"finish_reason":null}]}

470+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

467471

468− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

472+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

469473

470− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"description\": "}}]},"finish_reason":null}]}

471−

472− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"Create test"}}]},"finish_reason":null}]}

473−

474− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt with"}}]},"finish_reason":null}]}

475−

476− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" content '"}}]},"finish_reason":null}]}

477−

478− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello bash"}}]},"finish_reason":null}]}

479−

480− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"'\""}}]},"finish_reason":null}]}

481−

482− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

483−

484− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

485−

486− data: {"id":"chatcmpl-53fb4d1a2d0399cfc29f411b","object":"chat.completion.chunk","created":1778597313,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":711,"completion_tokens":61,"total_tokens":7812,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":10}}}

474+ data: {"id":"chatcmpl-452f541934d95f102a74b380","object":"chat.completion.chunk","created":1778601694,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7776,"completion_tokens":43,"total_tokens":7819,"completion_tokens_details":{"reasoning_tokens":42}}}

487475

488476 data: [DONE]

489477

@@ -492,15 +480,15 @@ interactions:

492480 - text/event-stream

493481 status: 200 OK

494482 code: 200

495− duration: 4.316007375s

483+ duration: 3.634534167s

496484- id: 2

497485 request:

498486 proto: HTTP/1.1

499487 proto_major: 1

500488 proto_minor: 1

501− content_length: 34150

489+ content_length: 34142

502490 host: ""

503− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'. do not print its timestamp\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_b10ffb78b6e844ce91c60f38\",\"function\":{\"arguments\":\"{\\\"command\\\": \\\"printf '%s' 'hello bash' > /tmp/crush-test/TestCoderAgent/glm-5.1/bash_tool/test.txt\\\", \\\"description\\\": \\\"Create test.txt with content 'hello bash'\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Create a file with echo and suppress timestamp printing.\"},{\"content\":\"no output\",\"tool_call_id\":\"call_b10ffb78b6e844ce91c60f38\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

491+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'. do not print its timestamp\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_1a51a6006cb9443db0b7bcb2\",\"function\":{\"arguments\":\"{\\\"command\\\": \\\"touch -t 197001010000 test.txt && echo 'hello bash' > test.txt\\\", \\\"description\\\": \\\"Create test.txt with content and zero timestamp\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"call_1a51a6006cb9443db0b7bcb2\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

504492 headers:

505493 Accept:

506494 - application/json

@@ -517,27 +505,45 @@ interactions:

517505 content_length: -1

518506 uncompressed: true

519507 body: |+

520− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

508+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

509+

510+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}]}

511+

512+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":". Created"},"finish_reason":null}]}

513+

514+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `test"},"finish_reason":null}]}

515+

516+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}]}

517+

518+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` with content"},"finish_reason":null}]}

519+

520+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

521+

522+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"hello bash"},"finish_reason":null}]}

523+

524+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` and timestamp"},"finish_reason":null}]}

525+

526+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" set"},"finish_reason":null}]}

521527

522− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}]}

528+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

523529

524− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":". Created"},"finish_reason":null}]}

530+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" epoch ("},"finish_reason":null}]}

525531

526− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `test"},"finish_reason":null}]}

532+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"197"},"finish_reason":null}]}

527533

528− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}]}

534+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"0-"},"finish_reason":null}]}

529535

530− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` with content"},"finish_reason":null}]}

536+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"01-01"},"finish_reason":null}]}

531537

532− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

538+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}]}

533539

534− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"hello bash"},"finish_reason":null}]}

540+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"00:"},"finish_reason":null}]}

535541

536− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`."},"finish_reason":null}]}

542+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"00)."},"finish_reason":null}]}

537543

538− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

544+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

539545

540− data: {"id":"chatcmpl-666738f4afd28160e089b9b7","object":"chat.completion.chunk","created":1778597325,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":768,"completion_tokens":15,"total_tokens":7823,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":14}}}

546+ data: {"id":"chatcmpl-590f5fe0db64d1bca1aad6e6","object":"chat.completion.chunk","created":1778601703,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7825,"completion_tokens":33,"total_tokens":7858,"completion_tokens_details":{"reasoning_tokens":32}}}

541547

542548 data: [DONE]

543549

@@ -546,4 +552,4 @@ interactions:

546552 - text/event-stream

547553 status: 200 OK

548554 code: 200

549− duration: 10.298101916s

555+ duration: 7.757482958s

internal/agent/testdata/TestCoderAgent/glm-5.1/download_tool.yaml+151−131

@@ -25,151 +25,171 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" asks"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summar"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ising"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" request"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" download"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" download"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" URL"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" save"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" https"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" as"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"://"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"example"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"-files"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".online"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"-con"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"vert"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".com"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/document"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/t"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"xt"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/example"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" save"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" it"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Possible"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" as"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Download"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" save"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" URL"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\n\n"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Count"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" length"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summarize"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Download"},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" could"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" save"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Download"},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Save"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" URL"},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".txt"},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ="},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" URL"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Download"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" and"},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" save"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" example"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" from"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" colon"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" URL"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

170+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" One"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-bd9d5eec2073c3b874b87534","object":"chat.completion.chunk","created":1778597326,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":34,"completion_tokens":80,"total_tokens":306,"prompt_tokens_details":{"cached_tokens":192}}}

172+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

173+

174+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

175+

176+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Download"},"finish_reason":null}]}

177+

178+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" and"},"finish_reason":null}]}

179+

180+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Save"},"finish_reason":null}]}

181+

182+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" example"},"finish_reason":null}]}

183+

184+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}]}

185+

186+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" from"},"finish_reason":null}]}

187+

188+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" URL"},"finish_reason":null}]}

189+

190+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

191+

192+ data: {"id":"chatcmpl-e4df7b056dc1c95101c2a8f0","object":"chat.completion.chunk","created":1778601704,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":34,"completion_tokens":90,"total_tokens":316,"prompt_tokens_details":{"cached_tokens":192}}}

173193

174194 data: [DONE]

175195

@@ -178,15 +198,15 @@ interactions:

178198 - text/event-stream

179199 status: 200 OK

180200 code: 200

181− duration: 650.606625ms

201+ duration: 467.600875ms

182202- id: 1

183203 request:

184204 proto: HTTP/1.1

185205 proto_major: 1

186206 proto_minor: 1

187− content_length: 33709

207+ content_length: 33796

188208 host: ""

189− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

209+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

190210 headers:

191211 Accept:

192212 - application/json

@@ -203,81 +223,85 @@ interactions:

203223 content_length: -1

204224 uncompressed: true

205225 body: |+

206− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

227+

228+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

207229

208− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

230+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

209231

210− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

232+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants to download"},"finish_reason":null}]}

211233

212− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me"},"finish_reason":null}]}

234+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" a file from"},"finish_reason":null}]}

213235

214− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to download a"},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" a URL"},"finish_reason":null}]}

215237

216− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" file from"},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" and save"},"finish_reason":null}]}

217239

218− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" a URL and"},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" it as"},"finish_reason":null}]}

219241

220− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" save it"},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" example.txt"},"finish_reason":null}]}

221243

222− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" as"},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":". I"},"finish_reason":null}]}

223245

224− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" example.txt"},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"'ll use the"},"finish_reason":null}]}

225247

226− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":". I"},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" download tool for"},"finish_reason":null}]}

227249

228− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"'ll"},"finish_reason":null}]}

250+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" this."},"finish_reason":null}]}

229251

230− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" use the download"},"finish_reason":null}]}

252+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_8e6a731f0e0a47a1b0017417","type":"function","index":0,"function":{"name":"download","arguments":""}}]},"finish_reason":null}]}

231253

232− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" tool."},"finish_reason":null}]}

254+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

233255

234− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_e0dedf0a68a44c4b82324ef5","type":"function","index":0,"function":{"name":"download","arguments":""}}]},"finish_reason":null}]}

256+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

235257

236− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

258+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

237259

238− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

239261

240− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": \"/tmp/c"}}]},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

241263

242− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

243265

244− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

245267

246− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

268+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

247269

248− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

270+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

249271

250− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

272+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

251273

252− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

274+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

253275

254− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"."}}]},"finish_reason":null}]}

276+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1"}}]},"finish_reason":null}]}

255277

256− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1/download_tool"}}]},"finish_reason":null}]}

278+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/download_tool"}}]},"finish_reason":null}]}

257279

258− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example.txt\""}}]},"finish_reason":null}]}

280+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example.txt\""}}]},"finish_reason":null}]}

259281

260− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", \"url\": "}}]},"finish_reason":null}]}

282+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

261283

262− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"https://example"}}]},"finish_reason":null}]}

284+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"url\": "}}]},"finish_reason":null}]}

263285

264− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-files"}}]},"finish_reason":null}]}

286+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"https://example"}}]},"finish_reason":null}]}

265287

266− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".online-"}}]},"finish_reason":null}]}

288+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-files"}}]},"finish_reason":null}]}

267289

268− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"convert.com/document"}}]},"finish_reason":null}]}

290+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".online-"}}]},"finish_reason":null}]}

269291

270− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/t"}}]},"finish_reason":null}]}

292+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"convert.com/document"}}]},"finish_reason":null}]}

271293

272− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"xt/example"}}]},"finish_reason":null}]}

294+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/t"}}]},"finish_reason":null}]}

273295

274− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt\""}}]},"finish_reason":null}]}

296+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"xt/example"}}]},"finish_reason":null}]}

275297

276− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

298+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt\""}}]},"finish_reason":null}]}

277299

278− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

300+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

279301

280− data: {"id":"chatcmpl-2f5df543a92ea1ec4fec95f9","object":"chat.completion.chunk","created":1778597330,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7754,"completion_tokens":71,"total_tokens":7825,"completion_tokens_details":{"reasoning_tokens":25}}}

302+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

303+

304+ data: {"id":"chatcmpl-1fc54652250a944af3fd063f","object":"chat.completion.chunk","created":1778601708,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7779,"completion_tokens":72,"total_tokens":7851,"completion_tokens_details":{"reasoning_tokens":26}}}

281305

282306 data: [DONE]

283307

@@ -286,7 +310,7 @@ interactions:

286310 - text/event-stream

287311 status: 200 OK

288312 code: 200

289− duration: 3.815614s

313+ duration: 4.006941625s

290314- id: 2

291315 request:

292316 proto: HTTP/1.1

@@ -332,15 +356,15 @@ interactions:

332356 - text/plain; charset=UTF-8

333357 status: 200 OK

334358 code: 200

335− duration: 1.644283791s

359+ duration: 678.656667ms

336360- id: 3

337361 request:

338362 proto: HTTP/1.1

339363 proto_major: 1

340364 proto_minor: 1

341− content_length: 34306

365+ content_length: 34399

342366 host: ""

343− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_e0dedf0a68a44c4b82324ef5\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/download_tool/example.txt\\\", \\\"url\\\": \\\"https://example-files.online-convert.com/document/txt/example.txt\\\"}\",\"name\":\"download\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to download a file from a URL and save it as example.txt. I'll use the download tool.\"},{\"content\":\"Successfully downloaded 2574 bytes to example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"call_e0dedf0a68a44c4b82324ef5\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

367+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_8e6a731f0e0a47a1b0017417\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/download_tool/example.txt\\\", \\\"url\\\": \\\"https://example-files.online-convert.com/document/txt/example.txt\\\"}\",\"name\":\"download\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants to download a file from a URL and save it as example.txt. I'll use the download tool for this.\"},{\"content\":\"Successfully downloaded 2574 bytes to example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"call_8e6a731f0e0a47a1b0017417\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

344368 headers:

345369 Accept:

346370 - application/json

@@ -357,25 +381,21 @@ interactions:

357381 content_length: -1

358382 uncompressed: true

359383 body: |+

360− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

361−

362− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Download"},"finish_reason":null}]}

363−

364− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ed to"},"finish_reason":null}]}

384+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

365385

366− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `example"},"finish_reason":null}]}

386+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Download"},"finish_reason":null}]}

367387

368− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}]}

388+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ed and"},"finish_reason":null}]}

369389

370− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` (25"},"finish_reason":null}]}

390+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" saved as `"},"finish_reason":null}]}

371391

372− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"74"},"finish_reason":null}]}

392+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"example.txt"},"finish_reason":null}]}

373393

374− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" bytes)."},"finish_reason":null}]}

394+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`."},"finish_reason":null}]}

375395

376− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

396+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

377397

378− data: {"id":"chatcmpl-a4f7722193027bafe61d2c96","object":"chat.completion.chunk","created":1778597336,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7826,"completion_tokens":15,"total_tokens":7841,"completion_tokens_details":{"reasoning_tokens":14}}}

398+ data: {"id":"chatcmpl-2fbaf1aacf59e1aafe55e635","object":"chat.completion.chunk","created":1778601713,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7851,"completion_tokens":11,"total_tokens":7862,"completion_tokens_details":{"reasoning_tokens":10}}}

379399

380400 data: [DONE]

381401

@@ -384,4 +404,4 @@ interactions:

384404 - text/event-stream

385405 status: 200 OK

386406 code: 200

387− duration: 3.863408666s

407+ duration: 4.159503041s

internal/agent/testdata/TestCoderAgent/glm-5.1/fetch_tool.yaml+334−311

@@ -25,367 +25,438 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" asks"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" fetch"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" fetch"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" https"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" https"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"://"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"://"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"..."},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"example"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"-files"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" tell"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".online"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" me"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"-con"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" if"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"vert"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" it"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".com"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" contains"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"website"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" word"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/html"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/example"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"John"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".html"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Doe"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" tell"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" me"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" They"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" if"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" want"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" it"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" contains"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" word"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" instruction"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"John"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Doe"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" colon"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Check"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" HTML"},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Check"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" if"},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"John"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".html"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Doe"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" contains"},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"John"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Doe"},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" could"},"finish_reason":null}]}

177177

178− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

178+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

179179

180− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

180+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

181181

182− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

182+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

183183

184− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Check"},"finish_reason":null}]}

184+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Check"},"finish_reason":null}]}

185185

186− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" example"},"finish_reason":null}]}

186+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

187187

188− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" HTML"},"finish_reason":null}]}

188+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

189189

190− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" for"},"finish_reason":null}]}

190+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"John"},"finish_reason":null}]}

191191

192− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" John"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Doe"},"finish_reason":null}]}

193193

194− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Doe"},"finish_reason":null}]}

194+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

195195

196− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" presence"},"finish_reason":null}]}

196+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

197197

198− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

198+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

199199

200− data: {"id":"chatcmpl-b39d5513375460df26bd477e","object":"chat.completion.chunk","created":1778597337,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":39,"completion_tokens":94,"total_tokens":325,"prompt_tokens_details":{"cached_tokens":192}}}

200+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".html"},"finish_reason":null}]}

201201

202− data: [DONE]

202+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

203203

204− headers:

205− Content-Type:

206− - text/event-stream

207− status: 200 OK

208− code: 200

209− duration: 497.991709ms

210−- id: 1

211− request:

212− proto: HTTP/1.1

213− proto_major: 1

214− proto_minor: 1

215− content_length: 33727

216− host: ""

217− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/fetch_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"fetch the content from https://example-files.online-convert.com/website/html/example.html and tell me if it contains the word 'John Doe'\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

218− headers:

219− Accept:

220− - application/json

221− Content-Type:

222− - application/json

223− User-Agent:

224− - Charm-Crush/devel (https://charm.land/crush)

225− url: https://hyper.charm.land/v1/chat/completions

226− method: POST

227− response:

228− proto: HTTP/2.0

229− proto_major: 2

230− proto_minor: 0

231− content_length: -1

232− uncompressed: true

233− body: |+

234− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

235205

236− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

237207

238− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

239209

240− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me to"},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

241211

242− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" fetch content"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Check"},"finish_reason":null}]}

243213

244− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" from a URL"},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

245215

246− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" and check"},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

247217

248− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" if it"},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"John"},"finish_reason":null}]}

249219

250− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" contains the"},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Doe"},"finish_reason":null}]}

251221

252− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" word '"},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

253223

254− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"John Doe"},"finish_reason":null}]}

224+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

255225

256− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"'. Let"},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" example"},"finish_reason":null}]}

257227

258− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me use"},"finish_reason":null}]}

228+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".html"},"finish_reason":null}]}

259229

260− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

230+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

261231

262− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" fetch tool"},"finish_reason":null}]}

232+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" length"},"finish_reason":null}]}

263233

264− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

234+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

265235

266− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_74e206507d5046e8a4b0aa5e","type":"function","index":0,"function":{"name":"fetch","arguments":""}}]},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

267237

268− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

269239

270− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

271241

272− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"format\": "}}]},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" C"},"finish_reason":null}]}

273243

274− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"text"}}]},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

275245

276− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

277247

278− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"url\": "}}]},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

279249

280− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"https://example"}}]},"finish_reason":null}]}

250+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"h"},"finish_reason":null}]}

281251

282− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-files"}}]},"finish_reason":null}]}

252+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

283253

284− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".online"}}]},"finish_reason":null}]}

254+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

285255

286− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}]}

256+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

287257

288− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"convert.com"}}]},"finish_reason":null}]}

258+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"c"},"finish_reason":null}]}

289259

290− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/website"}}]},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

291261

292− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/html"}}]},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"k"},"finish_reason":null}]}

293263

294− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example.html\""}}]},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

295265

296− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

297267

298− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

268+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

299269

300− data: {"id":"chatcmpl-74057d7e93cd8897f7724da7","object":"chat.completion.chunk","created":1778597340,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":721,"completion_tokens":57,"total_tokens":7818,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":28}}}

270+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

301271

302− data: [DONE]

272+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

303273

304− headers:

305− Content-Type:

306− - text/event-stream

307− status: 200 OK

308− code: 200

309− duration: 3.373965084s

310−- id: 2

311− request:

312− proto: HTTP/1.1

313− proto_major: 1

314− proto_minor: 1

315− content_length: 0

316− host: example-files.online-convert.com

317− headers:

318− User-Agent:

319− - crush/1.0

320− url: https://example-files.online-convert.com/website/html/example.html

321− method: GET

322− response:

323− proto: HTTP/2.0

324− proto_major: 2

325− proto_minor: 0

326− content_length: -1

327− uncompressed: true

328− body: |-

329− <!DOCTYPE html>

274+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"f"},"finish_reason":null}]}

330275

331− <html>

332− <body>

276+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

333277

334− <h1>HTML test file</h1>

278+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

335279

336− Purpose: Provide example of this file type

337− Document file type: HTML

338− Version: 1.0

339− Remark:

280+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

340281

341− <img align="right" style="height:5.0551cm;width:7.62cm;" src="data:image/*;base64,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

342− qIIqIIXEYw6iEq96h+m9Qq7V/T+nLffdxI6M3Ufk18+PVsk8Me0DqGNYLav/CnIqtVLFf8AGXbxF1vGd5O3nZXrTcUGPmM729RCuuV+lZb+pkOtZu4vXyQ2bZFzH/UvftU4d6Ke/Mitu+pAXkd0MgPNTdZZ5VjuwkDY5D9+LYi4/TCWNV5OSMlVzc7W8IholFqU77AvYZFtKnkO49g3zdTctdlQRscWTLzHxa/8tai78Bp9yinwlXCFq+VpUsSImoCSdaioTFHlt7bU6hW3+sbDJlmRbjlr2Mx+pVVCzNp7HUcmzmLBOk5r9izqLjqC5vm7MaZGU70NYXqxbTYL83ibbDGyvTHyFCvkJ3qbzq+4K9p/WevjRZYOOTVyr4cqnarhmXV0tY697uzkJSjlUt5WZOxav1FPFWfkressACqfIMxAWvrtFVwzGa1We3JsqLTAp+PLbbe/ZYr4r0pRkGytcagXdqtv08axVmX2aIulbJt50dXXt3+TCPXEUG5trH5WMq1SxTyHifQJ9juINR3mzNecjHSxlycRqUupvdu3W14rvlddGJis/wAe7ZrXAF9lWWwuyBfzXEZymZfwxabm2hIma6IL85Frw15UXUZIs7jXZF3Ki7CSqynKZUvVuXT6MuyWZhtbp96kmrlk3W1/E7hBKDtAAvaeSh9S0einU8mbVRX7Qzzx+oCBMNtPR27Ma+oLXjjvV912gv3LbBk4td1rjqLsl2J+tY3KtBtMZHPC7JfIGItaJ22uPXL6Tj7MP1jlCHtIlhZUbxOTT+xJM4+qCHyx+vuJ+3IsVOn1Khyc07ycnqrqcFbON+ZjXhqslH+5XkNL67GGI/tWrmvqj2CtbSkyXLWMifH3vFVuUD6Zq9Z2W17JTbwxWZrL8BmOM/crnd41dOt1bZfWca4gWMEMsI7eP4dm9YdifxswtuA+oPsdcdjiNEL5mOr/ABjyvr6bltUC6Ui7/wBrK1TU93CwdcNzW0r2q6eIuoZ6kx2TdtI+dTkLW+GeLf5QjV9YJ2S3musdlqnU2gGLWjGxPFeiN6h+qvobMO9rWda9fkpkovTwbMAVWtV05EFu6bs1lx689r676WRreKkUXnSbNmHbaBk9u2vMxjj3Ydym/OULj0hnGHZaxr02E5DZ2Rk96rpifIzOp1MnUMRlanM02Of+NSdVO0t4lljNK/LNowa4cgsc+zbcmDlr+x3B5Opjj9RS6xxxvtTkpsIAptSl1dRi9+58k9zMoUW3XOPkZyr8f5fEsbrMrExqkq6li5GNVba97JAhgLJO5ueSpFgD7aJ4mjtl4z6TH8knU2ODvuYB7B6Jfbk5nQhfyHEvm497XZvP49tvOp04rWW7dW+SuOVcZqsmzqALYGAbK8hXGVR00VVV0Y7r1PIf44W8VWZV/M4tvx1usN13zHWM/Kb5H6VN6dtHyYF3YFKv/wD1fWuQEc+eXod7s8MPLMvMcPdz5wkSlKLGpmTbj2hjwGOqvn2WdpHsmS7V2144GIx+Ph9Po7avqzH6o71HBLLZhvYZ/lOcyD+i8jKvexttLFQC3ucndzA/qnkudSshiQOFfgMPFSklyEOXRdXb0yxP9b0zvZGQ9zd/VijqaYl5zCDOfqmhRQgZhxLcuMwA75GboVo6LfXkcS1x7qXoLOoXAU5Dg2WkcteiHct/eG9f5niOTsERdQN48mMdr6kMfCk6rMsiLE3otqLvlZerirqSKLrBlytkanpdZdVNlb2K/Puta9Jr71pXWN4iOa16tmNm5WEBF6riKOs5xzbG8SsFmqbTDuaNmozOQ33E8HwYiDhOHOENKieNkx+oYYpo/wBfi14eSLWLWfJaotkZODlplZX6L8VYOAoT9xtYvW/mu1gj3rYjaFpvctZfZsWebbPVeOo2wE+rfMO9ACb1PGifYajs05GbjiINl040pPoMPU+BX4q/db/b+1myen2ObL6EtnTXZMU36dAd3v8A+9jDXUMsrThrkIcTqFIfpqVl3bIbn/Wxm2x3EOjUG5WdwzQEJJhZYFGuIMb6UeF8Hk29+qGWP4w8RfkZOOHsCLTjW8WxkyGFVfL4prFbV+968WezjXD7BB7VhQHFfaYrB7EleacgW/49iLvlxJprMb2dvDHfEctaOyfZ/Efy24dzxtPEuMBiAmKPaz3jECKAJs8FY6GjFRar6MjVqZPDJybPSm5uNitldRoqV+r9WrQv1Z/9d1DO602Zi2g10QeQ3hu2QyU8AjHk54qCxLfWp5m2nnaxRuftBMV/H84JYWJ8bFqx8db17drLZXdVXc+WjZFeTXOO1O4QzV8e3NiKAZeqqD4bG1ojZ8RnPFlHKz/kH7XAErQNLh7KNVp4WocpxE+7PZSs1FU8vEu8xKxyduM5+qnZbyd+iBnNHTtmvpOAyZnSV1jKVyDvu143O6qoYWZ0ptdRyLEweqrcz9R/yusNV02xExkppzem4+KXl1PamRh2XWfFeqqwxBG+j6xz42IzQaiMNosZhpmjNsoBCZZR8XERslp0xmswqudltDOLMay1q8+98q7mRWbF7NOiB++orE0strV0uqULXvjYIm4fsHccGaHBzsU6142SJuJ9nQdtbYxFHcrT35ebHIKaIZzp9z+F+z5iruY2P2pgebcsKuTbbwsztZy5aj5NWUK7czjemFZtuoVocZcg475WXfkRSZlXW0Nh98Gq8CLkbOUU+OTt38QbJc+XUmfU8kqjEhVWBvViY0VNzjocyCzaj4lDCyymvFGDbbfjmdWauzBbRPcFjBN2V+UsUqaGaKeUL6AbkR+FgXbJ+6w7jn9MRG/TLbLEa35rInL3U+5MB8b9d+PMUwnUaN4m/BPrhEi62znSlpFS39zHakXW9UruxK78kPioFe1a6K8OstXM8M2T1GzSIeErq4BKK1Tt9qKP1fHNjzHDuWX0PWFbcCy1wB66T7H0FJPEgFGJSgxaZkgKSV31K+ykZ+RzTFrsvfA3/qspuVnUDe+ay+rIRKG0ncbV9ZeukcYn7mUcCCjHZm9Dt7AAVDs37/VJhAnjt/2Jn9laLFaN+7fjfgk/gft8weYftfIb9tbCPcxnfLDBtX5D9RPxurdRN1Fa7x7f0Jm3gLj8Di24iFbKrLL7K0rlVzkm31tbRrrBCJZoOtcxu0g6tZxr2dKbOHEmJUTEpgr0FTU4TgBNTRmYbQ6kiZ/UsdcG7iTX8GvFGXVmjDwqsJck4mPS3F1RluK468k5pOQYgMJUVMLNu6L5jKVqrYBWbcQcTZoP/DfZ3+D6z+B9eeI1Cfx/XyI0b8eYRuU+I6lmNYmtsiK07VdcSwc+pn2qLHFs5FsZdZGwtHI1j9WwWVMqYm4rtvu7Deaq2tEQlovpZbgNfk24YocpBVAogEA/G59w/X0Mqo2IcV1BpMx6rP8AXV4jlOldK+NmWWKWzehvbLqcipK1NcbTulyva9Siysztpo+Ba3Kqr7sCkOuns/dz8nyP4/vNT+PsaGo37T4B+nHgbEP1oTX4fcTxUojsOJPtjkJFvbud1iLvdk7QovZRG8UZxauml+QO5a1lq0KFTyZjUnfG2IXptVtFiDYzskyjyv1+dTUImtn6hM3snxH1qrGq+Tjvj1UNkDGGPZkNdkA3Rse/jc+R3W9o9HFMVF3aRupwLQNTkONgCms8ST6E+7/SzctA3/Y/jxOO/wAHcx8ay5/iY9Q44/HPx1rA8D2jHyq+pPsda23b+hr1ABniEgTHavtWILKeya4/tRbezWZDlq01uhvKcSihSPpS5QhxP0ydVgr4t0BXcvGz8ctTluMST55MSR5hB15EO4fJs6dca8JESq2gPmW6WqvJuFld/Kdbx1YZCgWmwynirXgpAvNQgMrLLLa9z6g8H+eX4B8WRYdQ/ncr9nNr7yMqxIL7BPkbPjc8weRyWEoY59mPhuUr2JoGIIltcd+3Uba2uTzU/H5V6+gIU0EvD6oOKoG2dbJQ8QOMEp2BUQC7Lq0e3GED8ek5LOXIn64wKYU8CoGU2f8AtGlQbSqrjsMg5dLmVUc1z8psah1cRVbSLY7MjGtKtPWIVPD9kt2APJizxyB1LYPxs7g8QfVR4kOWWzW+2zRqbmgxbIMMmfDBi4dfIYlUGOgllHud8mPlmhEQOGtrDE1hsc1cTiqObclZU+RXXWKha3sG5OBshROPGC3iyKWbxz3HfY4sx8AtZ5AMIh8Q+YFnGaH4I3OE7VZNtnbpyM4319Kt3e+Oplo7M6xmY7hjp91MQS632jktntTrl3GE0tku8TwSymDTRvs72VLj4B7K4s7FcGPXs1CdkS1NDme5X+pKkGuIn8TUK/jUZC0fiiP5JPGFp/WtbJsGqvRGKxZ7SwyKWjrXar1ukRStU149tJUJR5HY4MW9mHJWq83UshSs7K6jRR51PWa8gfk/dPUOFrZKvYcU2PTQwsIAl27cTNQq2q7FCpusstl/6qnesd/LD2VdxgIzJyP2v7m+128q1UCuyFmhNTU0JmoQo8yj0QH3GzAJr8H6348fi30TIs7n4Jh8x+4Ex9ctKlON6M9jtkp9b9ydz2NI8EHxXYY9oLJ9VuTLOatyJgJMya3avzGXcUfjRMAXexs75RtQsZmDYw7Aa8KoVK6L28fq+MZ1jqPZx2stsTysRACieLe41ntNPWKn5oW0WO46vyYaixKu5K0CzMPvR7jiBAJxmh+MoBgmw6KxlK+Qs1CDNRiFNY8aH4zGUV2JXtvsH9MeQxaYRLJlMa7McmcFWwE9zXtK2O98o0JIAILC0CUWKZZd5L2CV2kxf+NwFgnIfhfv6J3NKRsTYnmJ3MiurHqZMAaqfqN9eR1O178h0d0C6DjQZ2RqlTXe8F1eIsCDS6MaLQHsC4QgwawwGgBMv/n6c+5ubE1OJhpJnx4uOgnAar9ciEGcYSJlH2QrqchM5laHGvaJh3avXtjn7dxpiWPXb1Otbciqg6rHclygWhRAsyKyDik8SPB+1X2VTrjxV11BuceMa1tc/B+gNfjYE5Q+RsfhjqeddTSvF6UMkb6zffVLMi5379pi3WArkssGWpFl9ZCWrOIEXFusPxbawJsA+Sa1JARgaVedpoUcH4/I9Oq4AVuVFR2ABNichOUJhMf/APY3DPcD9SZD/qV7AbZLJ5I8oqM1+QNXrfY/ZaW4TNKMN6VYebE7aOy0y9g5xCli8Jm1nhiLyrdNR6tRgUlGwNNFCgrxZXTlT2zO157R0KmE7LTstO0TDVqdudozs+ezM3p+Pm4WV/jTAN/jmSXu6LYhbpWQIcDKE+PYIKDDjbi4tgZcflKFtrnNtaDMEq2Ernr+Nzc2fyRAuppYNfn6hsQQ2tyDWT9TfGwjg0OOhgpUTtTtTtTtwouyiiJ6wPye3zUzHv3WbmW4ZQfTo9df4yNmrp/3b5i6Q7RjtYpXYHnflvrRgXYm/K7nifyTOcFimG4aFsbrHtV1AmLnJHyMa2GrAefBVp/ri0bpgh6f7f65t/68iN00z4yh/ijfxUnxfPxzyNMarUFeoVmjCPxqcfwfvx+CkKCBYVnjlYV2e1z7djEY8VdRgOKaI7YEZK9cRLe9y7eQtdbi+u81oBWHVRpel7GPsRtEYKlZkORC5i2tP1DBy3URsj25ahsm2M8zmRGs8h5y3AZ4355OFK1lTGbybRGZGa3jxAiWskGUS3y35/Lr0MuDKM+S7TlbGNgNjOYrcWLzkBN1zlXP04wTWpoCNP5M3+W5TnZzL2KLMljChsNOMeVePWs46YeDxG+IUqFhnHZK7nGPxWWBFtuocjGUGtxxbALfGOydOa8McqrNtO344ARLP15/YNCf0jyYMDy9o3svneiQQwhcgrszR0rMGcLCzzlbOVkaxp3n0bjGIed2C0Tkm0evTmucq4Hrn3CPULBsTW4RZDyi/RSaAh3yMP51PE+5pJ2lBux7WNbsrDJpndDQ6actEkzluE7mx+C25kULdX8GtExBbe70c4OXLpI5Y/Ebrs5ZYfjbw0ebkibIfxOJIpWeVgr8mkxUVSUhQAiceQ4JOCwKY1QM4kn+f54mcG35ELNAWIX8HxPG9/jepyWCA6inY3PE1DzWK02sH2xnOHZh5Qb/AAYdEI5A/iX41VjaZFDNFLTuPsspclYHaBmgYwGa84qkjjM1OGZ0hgrZ1q004ldddKpVvRY9pAbEBi/86/uX2iJ4/G/O2MMEZgINgA6mtQT12VM0NzztvpSVm/x/B3Dz0Nzc+/xr8eIIRGXg3IwQ6hHj6A0Zx3BuaP48w7m/JMdm0HmwQDqOpZvO3r5TsQK5l9WSIxsxwHcxL0MD8ozkBzYZnDUxWIvydXWLapWrgx5acbM/jGc94cWipFUJNzztuM5rvxGE8T1M1o+fwGM5RuUUHR+vEb1gc68kedQjxwJnBSERVniczOREbzEBn6m13rcGtDzOP49jGDb+oZ5Ijo5gVpwnHjGQMOLcdsDz87gnHcKeeBlitLVcl8ZTGouRa8h0C5HMZjBqcduMxl1HQFuyY6hjt1nd02I7hFLGWWEBbY1ilUZJ4hIDcjqu4tDZyRCIvgKZyn1CRGpMSsxajOB32LOS02bNNvEBu4VYTiNtXHrENYgA0RqGfwD+AZ9zXkLqanHwPoDU34biAIGMPLSMonIGE+PGrA5ZQ80YAZ/JIn8HuRlE4GfrrDXf3DVkS3FlioteIDDz5IwB7qxbQWJG8x1SmjJRK6ybRx5GuqVVoHtLcvaA7LbhHnfGG9CeSzuid7YazbbGg1Lzs0a7dU7aQVmMNQDaHt7sCQ64u4EVQ7tjsxWsQUAKUG7KKmD4w4mllIpEWpd9lQGoEVVjpxlrBZ3yZjOLCyeGqG9VwLsdkxq9Mp2G+gq6Ktvi2uB0RNDlpZx8keGTz2js1wpDVuHGBi43F3DbVPaynYC21rb8ucrGdXYWU3R8jUozW71+Qa6mzneJkjnW3g6Dhk4bVoRsabXDhNmFmWOLBP1FhyreSZFgYZLSrLnyhDmcm+UA7ZVUvtpMVqzLKqyi0MtJpvsnx7mjV3GV12CKjtLV4xW8PyVV3yyqn2BqGkkdpSPjr26scIpRw2+RSkPDjuk7T8ez7PjAwYq8Th6rqR1nbRwKE4fBWfCPL4y82wp8JtnGtllNnKyhkIx3MekzsGWUPOFomn0qMZ2WMso8DHUG/AQmrpw5WYPIt0z1HT3Qph2ayKTK+6JyuDG1wKchhFzFE71bTuVid1YxBDuQqUDhdhh0XEs43YorWnH9LEKHttzfFGrannaZ1aphBUQfqJsRWs1yujWWSxrd87I9lkay3QuteDIKlrixN7grkEz5Hhb9myxGHNIbmSLlyzK2q38ibWheyLa3BL3Ba+wk5T8a8gyvME+WpKWnmMu0lc4BbMtCoyKuPfqMZ6GZTTys+KZWmLxb4yvyx96x9BadcKp262laUa7aidtJbi45jYdXbfCv4NiZagrchtxxv4aPH6dduzDvQfqBVe0FMlxLchtaCNVapmiVX2hBjK7x15BFbYrBYKunrO3rLF6lWW0gsF4vUm3NMegmWUWRap29EVGdtyV5acvBZxZlJgQ9vg4h8BT4Ybg3OVZchwF7hNIbTBlbj4FZ2a5XVxKan1CSYDCDNe4sXhzHELsGkGK3kG0qVsE2YWE5QMYCdAx28QE7DusGRaJ32iZU7mNrhjGdigl8INLMA8r8ARsbGlqAxKhEBWIyzcZVn036YPJNBk3xVgyQ1OXaptGl2nx9Ds84aDOy4SpWKrTuWVaQ4+41REXal0O9ExFsgV+2Md2NeP7/AAhv4uocNImKhBxgJRUvcsSsscfxZSeTUkyyoqvHU4blg4zY06+P7kMsQv3NttKeddmItcHMXc4zT0M9NWBSFRYavH8lBvU7ZKcLAzqwm/HLibCzRbXRa77Fny3WfOsIbMsj2IYSgZjqbTWwFDAwmWJziVhE4ieJ/P8AJ5Qc43gBR+AfHiMEVdAzhGqWdhDDj+TSixVG+KwlQKyCPUz+PWBUK8RtaVVrKkMFQWGsRKxtqQScdNtQsuwuRTArC3Ybg04VgdqfZ8RouL5qx+DoocWYoY1YFqFMUOhx63luMhBwxwXC1Y2ITDhAPbhKwtwF7teJyl+ERHxbQDWdbBCLLBucTGExbRTZeA2WSrRUt3ch06eiqTLU3KxoQEmcvHKEnXOERjo1ly1g5L7ouzrKe5ImWchrMyyq6m4244ZzOQ2x1LLSK6rTvlxbQIX1O/cQ/gxWO/48b8M2vPvDORAWDg9ngLyjqlilBBWFD8N9yvXfEGSGPdVYMmrmxBDPpGcFE4y7nxM+Rcr12MbIF2bAK3sx6ngw61Awv00wuMvwmAbBfl8ezka+MA3Fn8/w3rE2JoleJ5HYJfVpLmOjmtCSuhrUA4rrc4+LGXXCmo8a7KuYrCupG9tHqFkFSQj2dwsdxFO0Phix0hYMfrfsGBDEGCKTtH2Wcw7ChvDuAe7uLYrkSh3J5tsHlOC9yyoPSuMssrMfHa2VVcFtrDVUVwV8X8cWHlRuW0WLaLSro6CPdQZXol6E5lG2oO1IsVjqcp//xAAoEQACAQQCAQQBBQEAAAAAAAAAARECEBIhIDEDEzBBUUAiMkJQYXH/2gAIAQMBAT8BpeoQ6IUv+6bg/wCezHCLf4NJMqa6X4kEe1H5NTg8jm65K88In/pA0l7EEEcYII/pW4tW/g37PxabIYhdWxd4II/pYtv2JMtlTlmWvZiyuuD0oGOv+odQtCYtjVk7MQ0dm1+GrLo/4JYr+oxEjES2NWiSBX6G0QjHi/eg6KpfY4+P6aBK8jZWpFSx9FOiqrZTWpGSKz2PT4LgxWavJNoG4P8ASf6GBuDITtA9IyKUR8E9xboT0LSln+lNJJF2pKuE8GuKGrvY4PjQ/wA2SSl/ZnvQnJ5CtSkU0w9WmBMj5KarUpJCGUjQ0kZCdmK3k6uhivNvgdkhido2NIUMdP5SX2dFQ5TE5cihaKtFOuh/qOyllWtlFbMkJlFUVGxLV/i1ZBT/AKLb4PY7PV4Iv8HZEDYrQM7E0iR2Y1H4EWkmENiRVBKHtiZXopZXVJIuiUxtEk6FbvZTuk3BDtJJVozKahVwZHZFvJF4u2J3m0XaO0VKGY2/7ZL7K/vlPJnV09Wq0PomTor7JhFFRMDUsVkioTjiikVquiRIqaR+5kQJbHTZDYzy2i1VQ6/o3ApFPyPm1I6btW6JEPs+B9GOpOyk+SLyhMqq3ZrVqlOiJWhjq4KlmWzu77PnkmU1EaEK6hWaNlIhklbtJW9HYoRkyljVkhWXCTK1UM6Q9WpHsmSgqqhCdmOtQUqUNMUtj0yjeyp7MtCqM5KGRLKoIu3N3d8o0hShuabIY2rriiu9QqRIRCt0SMm7OxKBjrJMvoh3xk+RHk6OmVM+B9aFb+RXJS9Wp6GLopZkoKrvg+/ZcxopKRqBC2NCOyIHpCIMXatbu3sVSEzJopq+B3qHbLjFmTeYtT9FRXTBQyvopUlDKeh/uKqpKFLPIv1FPQ7Pom3xzj2PG9nTKaSrezybKVA9lIiZZUKoStUmtjGSdmJpEoWh2kY7QJC0Ips0PRM8UaqKlKKDs8aSMf1QNwhOGVOSlwPbM3xfuTBIhU/YipfJJ/EqcFDtTvoqeyIGMpTZJXV8HzarSOhNigSQ6SIEPQhkjEh07J0K9VKY+xrjRMlQiSTLY2fIyT55Md1xYrMiyqFWiU+rfxIIKXCKXqRMdqqpZMIfkQ3NmNSOCBQKzNjdsoJGIyRKQnN3TU2OmHabpiZl92yJG9kz+AzLY7TZnwK6FVJ/E7NlKlHZBUypaPHTLK6pdkhIYxRbohjZNm7QrQQYkHXCqT4sh2Q4918lZ2jixWRXq1LhCezs+CikpiSpxVoqrlFWldEjqGrq+uMEWQzuzklyNffN2d4srtlJVZ65OyMhEEW+bLZgNaKlBT0KyKao0eN7PIUlSlDduieKRAlZqyEJDs+jfdurL/SqqebRUotI+T5xaRiHydqFDGVDclDtMDdn

343− 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

344− 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

345− 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

346− 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

347− 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

348− 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

349− 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

350− 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

351− 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

352− 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"/>

282+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"r"},"finish_reason":null}]}

353283

354− Example content:

355− The names "John Doe" for males, "Jane Doe" or "Jane Roe" for females, or "Jonnie Doe" and "Janie Doe" for children, or just "Doe" non-gender-specifically are used as placeholder names for a party whose true identity is unknown or must be withheld in a legal action, case, or discussion. The names are also used to refer to acorpse or hospital patient whose identity is unknown. This practice is widely used in the United States and Canada, but is rarely used in other English-speaking countries including the United Kingdom itself, from where the use of "John Doe" in a legal context originates. The names Joe Bloggs or John Smith are used in the UK instead, as well as in Australia and New Zealand.

284+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"9"},"finish_reason":null}]}

356285

357− John Doe is sometimes used to refer to a typical male in other contexts as well, in a similar manner to John Q. Public, known in Great Britain as Joe Public, John Smith or Joe Bloggs. For example, the first name listed on a form is often John Doe, along with a fictional address or other fictional information to provide an example of how to fill in the form. The name is also used frequently in popular culture, for example in the Frank Capra film Meet John Doe. John Doe was also the name of a 2002 American television series.

286+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

358287

359− Similarly, a child or baby whose identity is unknown may be referred to as Baby Doe. A notorious murder case in Kansas City, Missouri, referred to the baby victim as Precious Doe. Other unidentified female murder victims are Cali Doe and Princess Doe. Additional persons may be called James Doe, Judy Doe, etc. However, to avoid possible confusion, if two anonymous or unknown parties are cited in a specific case or action, the surnames Doe and Roe may be used simultaneously; for example, "John Doe v. Jane Roe". If several anonymous parties are referenced, they may simply be labelled John Doe #1, John Doe #2, etc. (the U.S. Operation Delego cited 21 (numbered) "John Doe"s) or labelled with other variants of Doe / Roe / Poe / etc. Other early alternatives such as John Stiles and Richard Miles are now rarely used, and Mary Major has been used in some American federal cases.

288+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

360289

361− <img style="height:1.6881cm;width:1.6881cm;" align="left" src="data:image/*;base64,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

362− B1/bMVO7nYqfTdlIhfMUvxUd51lhy67Tyq1UqJI9Jew1tnpUt0B6+/rwB89+fbtcyp1Q+gRk7z/Gy9tfzMHyrrfpKG8eqVDetsaC8o5Juz6zqvKuz6GyvKtkeXn3fbDzOauVdwpGynvzd+wv64P1QRgMBZhAGAwFmEAYDAWYQBgMBZhAGAwFmEAYDAWYQBgMBZhAGAwFmEAYDAWYQBgMBZhAGAwFmEAYDAWYQBgMBf4/fQGdtWECNsgAAAAASUVORK5CYII="/>

290+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"10"},"finish_reason":null}]}

363291

364− File created by <a href="https://www.online-convert.com" target="_blank">https://www.online-convert.com</a>

365− More example files: <a href="https://www.online-convert.com/file-type" target="_blank">https://www.online-convert.com/file-type</a>

366− Text of "Example content": <a href="https://en.wikipedia.org/wiki/John_Doe" target="_blank">Wikipedia</a>

367− License: <a href="https://creativecommons.org/licenses/by-sa/3.0/" target="_blank">Attribution-ShareAlike 3.0 Unported</a>

292+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")'"},"finish_reason":null}]}

368293

369− <img style="height:0.6949cm;width:1.9741cm;" align="right" src="data:image/*;base64,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"/>

294+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"11"},"finish_reason":null}]}

370295

371− Feel free to use and share the file according to license above.

296+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"J"},"finish_reason":null}]}

297+

298+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"12"},"finish_reason":null}]}

299+

300+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

301+

302+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"13"},"finish_reason":null}]}

303+

304+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"h"},"finish_reason":null}]}

305+

306+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"14"},"finish_reason":null}]}

307+

308+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"n"},"finish_reason":null}]}

309+

310+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"15"},"finish_reason":null}]}

311+

312+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

313+

314+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

315+

316+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"16"},"finish_reason":null}]}

317+

318+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"D"},"finish_reason":null}]}

319+

320+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

321+

322+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

323+

324+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"18"},"finish_reason":null}]}

325+

326+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

327+

328+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

329+

330+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"19"},"finish_reason":null}]}

331+

332+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

333+

334+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

335+

336+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"20"},"finish_reason":null}]}

337+

338+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")i"},"finish_reason":null}]}

339+

340+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"21"},"finish_reason":null}]}

341+

342+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"n"},"finish_reason":null}]}

343+

344+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"22"},"finish_reason":null}]}

345+

346+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

347+

348+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

349+

350+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"23"},"finish_reason":null}]}

351+

352+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")e"},"finish_reason":null}]}

353+

354+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"24"},"finish_reason":null}]}

355+

356+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"x"},"finish_reason":null}]}

357+

358+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"25"},"finish_reason":null}]}

359+

360+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"a"},"finish_reason":null}]}

361+

362+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"26"},"finish_reason":null}]}

363+

364+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"m"},"finish_reason":null}]}

365+

366+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"27"},"finish_reason":null}]}

367+

368+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"p"},"finish_reason":null}]}

369+

370+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"28"},"finish_reason":null}]}

371+

372+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"l"},"finish_reason":null}]}

373+

374+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"29"},"finish_reason":null}]}

375+

376+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

377+

378+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"30"},"finish_reason":null}]}

379+

380+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

381+

382+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"31"},"finish_reason":null}]}

383+

384+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"h"},"finish_reason":null}]}

385+

386+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"32"},"finish_reason":null}]}

387+

388+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"t"},"finish_reason":null}]}

389+

390+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

391+

392+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"m"},"finish_reason":null}]}

393+

394+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"34"},"finish_reason":null}]}

395+

396+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"l"},"finish_reason":null}]}

397+

398+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"35"},"finish_reason":null}]}

399+

400+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

401+

402+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

403+

404+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

405+

406+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"35"},"finish_reason":null}]}

407+

408+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

409+

410+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

411+

412+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

413+

414+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

415+

416+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Provide"},"finish_reason":null}]}

417+

418+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

419+

420+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

421+

422+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Check"},"finish_reason":null}]}

423+

424+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" for"},"finish_reason":null}]}

425+

426+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" '"},"finish_reason":null}]}

427+

428+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"John"},"finish_reason":null}]}

429+

430+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Doe"},"finish_reason":null}]}

431+

432+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"'"},"finish_reason":null}]}

433+

434+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

435+

436+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" example"},"finish_reason":null}]}

437+

438+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".html"},"finish_reason":null}]}

439+

440+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

441+

442+ data: {"id":"chatcmpl-f0927d49937b379d557ec87f","object":"chat.completion.chunk","created":1778601714,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":167,"completion_tokens":215,"total_tokens":446,"prompt_tokens_details":{"cached_tokens":64}}}

443+

444+ data: [DONE]

372445

373− </body>

374− </html>

375446 headers:

376447 Content-Type:

377− - text/html; charset=UTF-8

448+ - text/event-stream

378449 status: 200 OK

379450 code: 200

380− duration: 165.532958ms

381−- id: 3

451+ duration: 597.545416ms

452+- id: 1

382453 request:

383454 proto: HTTP/1.1

384455 proto_major: 1

385456 proto_minor: 1

386− content_length: 34286

457+ content_length: 33814

387458 host: ""

388− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/fetch_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"fetch the content from https://example-files.online-convert.com/website/html/example.html and tell me if it contains the word 'John Doe'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_74e206507d5046e8a4b0aa5e\",\"function\":{\"arguments\":\"{\\\"format\\\": \\\"text\\\", \\\"url\\\": \\\"https://example-files.online-convert.com/website/html/example.html\\\"}\",\"name\":\"fetch\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to fetch content from a URL and check if it contains the word 'John Doe'. Let me use the fetch tool.\"},{\"content\":\"HTML test file Purpose: Provide example of this file type Document file type: HTML Version: 1.0 Remark:\",\"tool_call_id\":\"call_74e206507d5046e8a4b0aa5e\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

459+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/fetch_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"fetch the content from https://example-files.online-convert.com/website/html/example.html and tell me if it contains the word 'John Doe'\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

389460 headers:

390461 Accept:

391462 - application/json

@@ -402,83 +473,61 @@ interactions:

402473 content_length: -1

403474 uncompressed: true

404475 body: |+

405− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

406−

407− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

408−

409− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" fetched"},"finish_reason":null}]}

476+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

410477

411− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" content doesn"},"finish_reason":null}]}

478+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

412479

413− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"'t seem"},"finish_reason":null}]}

480+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

414481

415− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to contain"},"finish_reason":null}]}

482+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me to fetch"},"finish_reason":null}]}

416483

417− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

484+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

418485

419− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"John Doe"},"finish_reason":null}]}

486+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" from a URL and"},"finish_reason":null}]}

420487

421− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"\". Let me check"},"finish_reason":null}]}

488+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" check"},"finish_reason":null}]}

422489

423− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" more"},"finish_reason":null}]}

490+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" if it"},"finish_reason":null}]}

424491

425− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" carefully -"},"finish_reason":null}]}

492+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" contains the word '"},"finish_reason":null}]}

426493

427− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the output"},"finish_reason":null}]}

494+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"John Doe"},"finish_reason":null}]}

428495

429− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" seems truncated"},"finish_reason":null}]}

496+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"'."},"finish_reason":null}]}

430497

431− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":". Let"},"finish_reason":null}]}

498+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_ee068c7ab33d4d7ca7398937","type":"function","index":0,"function":{"name":"fetch","arguments":""}}]},"finish_reason":null}]}

432499

433− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me try"},"finish_reason":null}]}

500+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

434501

435− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" fetching"},"finish_reason":null}]}

502+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

436503

437− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" as"},"finish_reason":null}]}

504+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"format\": "}}]},"finish_reason":null}]}

438505

439− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" html"},"finish_reason":null}]}

506+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"text\""}}]},"finish_reason":null}]}

440507

441− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" format"},"finish_reason":null}]}

508+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", \"url\": "}}]},"finish_reason":null}]}

442509

443− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

510+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"https://example"}}]},"finish_reason":null}]}

444511

445− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" get the"},"finish_reason":null}]}

512+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-files"}}]},"finish_reason":null}]}

446513

447− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" full content."},"finish_reason":null}]}

514+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".online"}}]},"finish_reason":null}]}

448515

449− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_3be36893074d462db576bd59","type":"function","index":0,"function":{"name":"fetch","arguments":""}}]},"finish_reason":null}]}

516+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-convert.com"}}]},"finish_reason":null}]}

450517

451− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

518+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/website"}}]},"finish_reason":null}]}

452519

453− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

520+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/html"}}]},"finish_reason":null}]}

454521

455− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"format\": "}}]},"finish_reason":null}]}

522+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example.html"}}]},"finish_reason":null}]}

456523

457− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"html\", "}}]},"finish_reason":null}]}

524+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

458525

459− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"url\": "}}]},"finish_reason":null}]}

526+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

460527

461− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"https://example"}}]},"finish_reason":null}]}

528+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

462529

463− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-files"}}]},"finish_reason":null}]}

464−

465− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".online-"}}]},"finish_reason":null}]}

466−

467− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"convert.com/"}}]},"finish_reason":null}]}

468−

469− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"website/html"}}]},"finish_reason":null}]}

470−

471− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example"}}]},"finish_reason":null}]}

472−

473− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".html"}}]},"finish_reason":null}]}

474−

475− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

476−

477− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

478−

479− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

480−

481− data: {"id":"chatcmpl-21dd60a7cd71549e9e6231dc","object":"chat.completion.chunk","created":1778597345,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":778,"completion_tokens":65,"total_tokens":7883,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":36}}}

530+ data: {"id":"chatcmpl-0367ea5f9721db65227d2560","object":"chat.completion.chunk","created":1778601718,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7786,"completion_tokens":50,"total_tokens":7836,"completion_tokens_details":{"reasoning_tokens":21}}}

482531

483532 data: [DONE]

484533

@@ -487,8 +536,8 @@ interactions:

487536 - text/event-stream

488537 status: 200 OK

489538 code: 200

490− duration: 3.695552042s

491−- id: 4

539+ duration: 5.000877791s

540+- id: 2

492541 request:

493542 proto: HTTP/1.1

494543 proto_major: 1

@@ -558,15 +607,15 @@ interactions:

558607 - text/html; charset=UTF-8

559608 status: 200 OK

560609 code: 200

561− duration: 154.72225ms

562−- id: 5

610+ duration: 158.890875ms

611+- id: 3

563612 request:

564613 proto: HTTP/1.1

565614 proto_major: 1

566615 proto_minor: 1

567− content_length: 34992

616+ content_length: 34346

568617 host: ""

569− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/fetch_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"fetch the content from https://example-files.online-convert.com/website/html/example.html and tell me if it contains the word 'John Doe'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_74e206507d5046e8a4b0aa5e\",\"function\":{\"arguments\":\"{\\\"format\\\": \\\"text\\\", \\\"url\\\": \\\"https://example-files.online-convert.com/website/html/example.html\\\"}\",\"name\":\"fetch\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to fetch content from a URL and check if it contains the word 'John Doe'. Let me use the fetch tool.\"},{\"content\":\"HTML test file Purpose: Provide example of this file type Document file type: HTML Version: 1.0 Remark:\",\"tool_call_id\":\"call_74e206507d5046e8a4b0aa5e\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_3be36893074d462db576bd59\",\"function\":{\"arguments\":\"{\\\"format\\\": \\\"html\\\", \\\"url\\\": \\\"https://example-files.online-convert.com/website/html/example.html\\\"}\",\"name\":\"fetch\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The fetched content doesn't seem to contain \\\"John Doe\\\". Let me check more carefully - the output seems truncated. Let me try fetching as html format to get the full content.\"},{\"content\":\"<html>\\n<body>\\n\\n\\n\\t\\t<h1>HTML test file</h1>\\n\\nPurpose: Provide example of this file type \\nDocument file type: HTML \\nVersion: 1.0 \\nRemark:\\n\\n\\n</body>\\n</html>\",\"tool_call_id\":\"call_3be36893074d462db576bd59\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

618+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/fetch_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"fetch the content from https://example-files.online-convert.com/website/html/example.html and tell me if it contains the word 'John Doe'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_ee068c7ab33d4d7ca7398937\",\"function\":{\"arguments\":\"{\\\"format\\\": \\\"text\\\", \\\"url\\\": \\\"https://example-files.online-convert.com/website/html/example.html\\\"}\",\"name\":\"fetch\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to fetch content from a URL and check if it contains the word 'John Doe'.\"},{\"content\":\"HTML test file Purpose: Provide example of this file type Document file type: HTML Version: 1.0 Remark:\",\"tool_call_id\":\"call_ee068c7ab33d4d7ca7398937\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

570619 headers:

571620 Accept:

572621 - application/json

@@ -583,53 +632,27 @@ interactions:

583632 content_length: -1

584633 uncompressed: true

585634 body: |+

586− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

587−

588− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"No"},"finish_reason":null}]}

589−

590− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":", the"},"finish_reason":null}]}

591−

592− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" page does"},"finish_reason":null}]}

593−

594− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" not contain"},"finish_reason":null}]}

595−

596− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" the word"},"finish_reason":null}]}

597−

598− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" \"John"},"finish_reason":null}]}

599−

600− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Doe\"."},"finish_reason":null}]}

601−

602− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Its content"},"finish_reason":null}]}

603−

604− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" is"},"finish_reason":null}]}

605−

606− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" just a"},"finish_reason":null}]}

607−

608− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" basic"},"finish_reason":null}]}

609−

610− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" HTML test"},"finish_reason":null}]}

611−

612− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" file"},"finish_reason":null}]}

635+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

613636

614− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" with a"},"finish_reason":null}]}

637+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"The"},"finish_reason":null}]}

615638

616− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" heading and"},"finish_reason":null}]}

639+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" fetched"},"finish_reason":null}]}

617640

618− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" some"},"finish_reason":null}]}

641+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" content does"},"finish_reason":null}]}

619642

620− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" metadata ("},"finish_reason":null}]}

643+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" **"},"finish_reason":null}]}

621644

622− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"purpose"},"finish_reason":null}]}

645+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"not** contain"},"finish_reason":null}]}

623646

624− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":", file"},"finish_reason":null}]}

647+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" the word \""},"finish_reason":null}]}

625648

626− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" type, version"},"finish_reason":null}]}

649+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"John Doe"},"finish_reason":null}]}

627650

628− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":")."},"finish_reason":null}]}

651+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"\"."},"finish_reason":null}]}

629652

630− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

653+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

631654

632− data: {"id":"chatcmpl-6756b8e9c05e7811b40c247a","object":"chat.completion.chunk","created":1778597353,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":868,"completion_tokens":38,"total_tokens":7946,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":37}}}

655+ data: {"id":"chatcmpl-2f895e7d8be2f4beffa3c809","object":"chat.completion.chunk","created":1778601723,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7843,"completion_tokens":16,"total_tokens":7859,"completion_tokens_details":{"reasoning_tokens":15}}}

633656

634657 data: [DONE]

635658

@@ -638,4 +661,4 @@ interactions:

638661 - text/event-stream

639662 status: 200 OK

640663 code: 200

641− duration: 6.389163583s

664+ duration: 2.902633375s

internal/agent/testdata/TestCoderAgent/glm-5.1/glob_tool.yaml+159−121

@@ -25,155 +25,209 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ask"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" use"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" all"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" all"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" current"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" current"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" must"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" produce"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summarize"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Something"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" like"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Finding"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Files"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Find"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" all"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Glob"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Current"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Directory"},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Maybe"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Using"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" output"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" locate"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Finding"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" ."},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Files"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" with"},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Glob"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Current"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Using"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Directory"},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

174+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-0c77c926525c023d5e9f33e3","object":"chat.completion.chunk","created":1778597355,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":163,"completion_tokens":82,"total_tokens":293,"prompt_tokens_details":{"cached_tokens":48}}}

176+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" locate"},"finish_reason":null}]}

177+

178+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

179+

180+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

181+

182+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

183+

184+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

185+

186+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ="},"finish_reason":null}]}

187+

188+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

189+

190+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"30"},"finish_reason":null}]}

191+

192+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

193+

194+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

195+

196+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" OK"},"finish_reason":null}]}

197+

198+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

199+

200+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

201+

202+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" as"},"finish_reason":null}]}

203+

204+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" single"},"finish_reason":null}]}

205+

206+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

207+

208+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

209+

210+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

211+

212+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

213+

214+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Using"},"finish_reason":null}]}

215+

216+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" glob"},"finish_reason":null}]}

217+

218+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

219+

220+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" locate"},"finish_reason":null}]}

221+

222+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" ."},"finish_reason":null}]}

223+

224+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}]}

225+

226+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" files"},"finish_reason":null}]}

227+

228+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

229+

230+ data: {"id":"chatcmpl-832c23734a41e749ff8db8ed","object":"chat.completion.chunk","created":1778601724,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":23,"completion_tokens":109,"total_tokens":324,"prompt_tokens_details":{"cached_tokens":192}}}

177231

178232 data: [DONE]

179233

@@ -182,15 +236,15 @@ interactions:

182236 - text/event-stream

183237 status: 200 OK

184238 code: 200

185− duration: 539.286125ms

239+ duration: 445.594917ms

186240- id: 1

187241 request:

188242 proto: HTTP/1.1

189243 proto_major: 1

190244 proto_minor: 1

191− content_length: 33645

245+ content_length: 33732

192246 host: ""

193− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

247+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

194248 headers:

195249 Accept:

196250 - application/json

@@ -207,47 +261,33 @@ interactions:

207261 content_length: -1

208262 uncompressed: true

209263 body: |+

210− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

211265

212− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_678a92c4217946c3a5ebf102","type":"function","index":0,"function":{"name":"glob","arguments":""}}]},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Simple"},"finish_reason":null}]}

213267

214− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

268+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

215269

216− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

270+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" request"},"finish_reason":null}]}

217271

218− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"path\": "}}]},"finish_reason":null}]}

272+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

219273

220− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

274+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_93398c73218c40abab6ccf80","type":"function","index":0,"function":{"name":"glob","arguments":""}}]},"finish_reason":null}]}

221275

222− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

276+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

223277

224− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

278+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

225279

226− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

280+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

227281

228− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

282+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"**/*."}}]},"finish_reason":null}]}

229283

230− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

284+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go\""}}]},"finish_reason":null}]}

231285

232− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

286+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

233287

234− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1"}}]},"finish_reason":null}]}

288+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

235289

236− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glob_tool"}}]},"finish_reason":null}]}

237−

238− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

239−

240− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

241−

242− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"*."}}]},"finish_reason":null}]}

243−

244− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go\""}}]},"finish_reason":null}]}

245−

246− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

247−

248− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

249−

250− data: {"id":"chatcmpl-1beb7bb732a1dc9c9224afe9","object":"chat.completion.chunk","created":1778597357,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7744,"completion_tokens":33,"total_tokens":7777,"completion_tokens_details":{"reasoning_tokens":32}}}

290+ data: {"id":"chatcmpl-ff98cd358f55061c688302dd","object":"chat.completion.chunk","created":1778601726,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7769,"completion_tokens":17,"total_tokens":7786,"completion_tokens_details":{"reasoning_tokens":4}}}

251291

252292 data: [DONE]

253293

@@ -256,15 +296,15 @@ interactions:

256296 - text/event-stream

257297 status: 200 OK

258298 code: 200

259− duration: 2.416067083s

299+ duration: 3.008176292s

260300- id: 2

261301 request:

262302 proto: HTTP/1.1

263303 proto_major: 1

264304 proto_minor: 1

265− content_length: 33999

305+ content_length: 34068

266306 host: ""

267− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_678a92c4217946c3a5ebf102\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/glob_tool\\\", \\\"pattern\\\": \\\"*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/glob_tool/main.go\",\"tool_call_id\":\"call_678a92c4217946c3a5ebf102\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

307+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_93398c73218c40abab6ccf80\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Simple glob request.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/glob_tool/main.go\",\"tool_call_id\":\"call_93398c73218c40abab6ccf80\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

268308 headers:

269309 Accept:

270310 - application/json

@@ -281,25 +321,23 @@ interactions:

281321 content_length: -1

282322 uncompressed: true

283323 body: |+

284− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

285−

286− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Found"},"finish_reason":null}]}

324+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

287325

288− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}]}

326+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Found"},"finish_reason":null}]}

289327

290− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"1 `."},"finish_reason":null}]}

328+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}]}

291329

292− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}]}

330+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"1 `."},"finish_reason":null}]}

293331

294− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` file"},"finish_reason":null}]}

332+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"go`"},"finish_reason":null}]}

295333

296− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":": `main"},"finish_reason":null}]}

334+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" file: `main"},"finish_reason":null}]}

297335

298− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go`"},"finish_reason":null}]}

336+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go`"},"finish_reason":null}]}

299337

300− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

338+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

301339

302− data: {"id":"chatcmpl-05f41bca68c4196b751cb426","object":"chat.completion.chunk","created":1778597364,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":759,"completion_tokens":14,"total_tokens":7813,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":13}}}

340+ data: {"id":"chatcmpl-b9e97a652a8e2fc7b0993d7f","object":"chat.completion.chunk","created":1778601730,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7804,"completion_tokens":14,"total_tokens":7818,"completion_tokens_details":{"reasoning_tokens":13}}}

303341

304342 data: [DONE]

305343

@@ -308,4 +346,4 @@ interactions:

308346 - text/event-stream

309347 status: 200 OK

310348 code: 200

311− duration: 6.472530833s

349+ duration: 3.883457292s

internal/agent/testdata/TestCoderAgent/glm-5.1/grep_tool.yaml+139−195

@@ -25,237 +25,157 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" grep"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" search"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" word"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"package"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" go"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" produce"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"using"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" grep"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"package"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" maybe"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"G"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"rep"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"package"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" something"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" like"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" source"},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Using"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" grep"},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" package"},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"G"},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"rep"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" for"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" “"},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"package"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"”"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Go"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" source"},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Using"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" files"},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" grep"},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

177177

178− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

179−

180− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

181−

182− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" package"},"finish_reason":null}]}

183−

184− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

185−

186− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

187−

188− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

189−

190− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

191−

192− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ="},"finish_reason":null}]}

193−

194− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

195−

196− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

197−

198− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

199−

200− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

201−

202− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Or"},"finish_reason":null}]}

203−

204− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

205−

206− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"grep"},"finish_reason":null}]}

207−

208− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

209−

210− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

211−

212− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"package"},"finish_reason":null}]}

213−

214− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

215−

216− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

217−

218− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

219−

220− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" source"},"finish_reason":null}]}

221−

222− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

223−

224− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

225−

226− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

227−

228− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

229−

230− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

231−

232− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

233−

234− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Provide"},"finish_reason":null}]}

235−

236− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" as"},"finish_reason":null}]}

237−

238− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

239−

240− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

241−

242− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"grep"},"finish_reason":null}]}

243−

244− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" for"},"finish_reason":null}]}

245−

246− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" package"},"finish_reason":null}]}

247−

248− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

249−

250− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Go"},"finish_reason":null}]}

251−

252− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" source"},"finish_reason":null}]}

253−

254− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" files"},"finish_reason":null}]}

255−

256− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

257−

258− data: {"id":"chatcmpl-20d39cc88cdd5a5f8d186ad0","object":"chat.completion.chunk","created":1778597364,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":24,"completion_tokens":123,"total_tokens":339,"prompt_tokens_details":{"cached_tokens":192}}}

178+ data: {"id":"chatcmpl-dd995ea66dc2661213ac8d4b","object":"chat.completion.chunk","created":1778601731,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":24,"completion_tokens":83,"total_tokens":299,"prompt_tokens_details":{"cached_tokens":192}}}

259179

260180 data: [DONE]

261181

@@ -264,15 +184,15 @@ interactions:

264184 - text/event-stream

265185 status: 200 OK

266186 code: 200

267− duration: 552.038958ms

187+ duration: 428.985583ms

268188- id: 1

269189 request:

270190 proto: HTTP/1.1

271191 proto_major: 1

272192 proto_minor: 1

273− content_length: 33643

193+ content_length: 33730

274194 host: ""

275− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

195+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

276196 headers:

277197 Accept:

278198 - application/json

@@ -289,43 +209,69 @@ interactions:

289209 content_length: -1

290210 uncompressed: true

291211 body: |+

292− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

213+

214+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Search"},"finish_reason":null}]}

293215

294− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

295217

296− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" 'package"},"finish_reason":null}]}

297219

298− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants to search for"},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"' in Go"},"finish_reason":null}]}

299221

300− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the word"},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" files in"},"finish_reason":null}]}

301223

302− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" 'package"},"finish_reason":null}]}

224+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the current"},"finish_reason":null}]}

303225

304− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"' in Go"},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

305227

306− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" files."},"finish_reason":null}]}

228+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

307229

308− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_596727092909456da0b8340d","type":"function","index":0,"function":{"name":"grep","arguments":""}}]},"finish_reason":null}]}

230+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_6e6529e4c8054555bcb2df8d","type":"function","index":0,"function":{"name":"grep","arguments":""}}]},"finish_reason":null}]}

309231

310− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

232+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

311233

312− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

234+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

313235

314− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"include\": \"*."}}]},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"include\": "}}]},"finish_reason":null}]}

315237

316− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go\""}}]},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"*.go"}}]},"finish_reason":null}]}

317239

318− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

319241

320− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"path\": \"/tmp"}}]},"finish_reason":null}]}

321243

322− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"package\""}}]},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

323245

324− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

325247

326− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

248+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

327249

328− data: {"id":"chatcmpl-8c7c12d18b49b0b51650ea6a","object":"chat.completion.chunk","created":1778597368,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7745,"completion_tokens":33,"total_tokens":7778,"completion_tokens_details":{"reasoning_tokens":15}}}

250+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

251+

252+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

253+

254+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

255+

256+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

257+

258+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5."}}]},"finish_reason":null}]}

259+

260+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1/g"}}]},"finish_reason":null}]}

261+

262+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rep_tool"}}]},"finish_reason":null}]}

263+

264+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

265+

266+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

267+

268+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"package\""}}]},"finish_reason":null}]}

269+

270+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

271+

272+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

273+

274+ data: {"id":"chatcmpl-de353447d4e7480858e0d2a1","object":"chat.completion.chunk","created":1778601733,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":730,"completion_tokens":52,"total_tokens":7822,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":13}}}

329275

330276 data: [DONE]

331277

@@ -334,15 +280,15 @@ interactions:

334280 - text/event-stream

335281 status: 200 OK

336282 code: 200

337− duration: 4.030870625s

283+ duration: 2.815924666s

338284- id: 2

339285 request:

340286 proto: HTTP/1.1

341287 proto_major: 1

342288 proto_minor: 1

343− content_length: 34094

289+ content_length: 34243

344290 host: ""

345− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_596727092909456da0b8340d\",\"function\":{\"arguments\":\"{\\\"include\\\": \\\"*.go\\\", \\\"pattern\\\": \\\"package\\\"}\",\"name\":\"grep\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants to search for the word 'package' in Go files.\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/glm-5.1/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"call_596727092909456da0b8340d\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

291+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_6e6529e4c8054555bcb2df8d\",\"function\":{\"arguments\":\"{\\\"include\\\": \\\"*.go\\\", \\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/grep_tool\\\", \\\"pattern\\\": \\\"package\\\"}\",\"name\":\"grep\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Search for 'package' in Go files in the current directory.\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/glm-5.1/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"call_6e6529e4c8054555bcb2df8d\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

346292 headers:

347293 Accept:

348294 - application/json

@@ -359,29 +305,27 @@ interactions:

359305 content_length: -1

360306 uncompressed: true

361307 body: |+

362− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

363−

364− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Found"},"finish_reason":null}]}

308+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

365309

366− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}]}

310+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Found"},"finish_reason":null}]}

367311

368− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"1 match"},"finish_reason":null}]}

312+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 1"},"finish_reason":null}]}

369313

370− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":": `"},"finish_reason":null}]}

314+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" match in"},"finish_reason":null}]}

371315

372− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main.go"},"finish_reason":null}]}

316+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

373317

374− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":1"},"finish_reason":null}]}

318+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main.go:1"},"finish_reason":null}]}

375319

376− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` — `"},"finish_reason":null}]}

320+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` — `"},"finish_reason":null}]}

377321

378− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"package main"},"finish_reason":null}]}

322+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"package"},"finish_reason":null}]}

379323

380− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`"},"finish_reason":null}]}

324+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" main`."},"finish_reason":null}]}

381325

382− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

326+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

383327

384− data: {"id":"chatcmpl-150a68d7d0d9db901807554d","object":"chat.completion.chunk","created":1778597372,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":763,"completion_tokens":18,"total_tokens":7821,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":17}}}

328+ data: {"id":"chatcmpl-ab3020d8b997a66d8be305a7","object":"chat.completion.chunk","created":1778601736,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":105,"completion_tokens":18,"total_tokens":7867,"prompt_tokens_details":{"cached_tokens":7744},"completion_tokens_details":{"reasoning_tokens":17}}}

385329

386330 data: [DONE]

387331

@@ -390,4 +334,4 @@ interactions:

390334 - text/event-stream

391335 status: 200 OK

392336 code: 200

393− duration: 3.065027833s

337+ duration: 1.829357958s

internal/agent/testdata/TestCoderAgent/glm-5.1/ls_tool.yaml+114−184

@@ -25,231 +25,137 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" system"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" asks"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" says"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"You"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" will"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" generate"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" based"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" on"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" use"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" first"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ls"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" message"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" list"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" begins"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" conversation"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" current"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\""},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" We"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" output"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" The"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user's"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" message"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Max"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Something"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" like"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Listing"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Files"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ls"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ls"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" list"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" current"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" just"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"\n\n"},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"They"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" want"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Listing"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Files"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" with"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" ls"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" less"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

157157

158− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" than"},"finish_reason":null}]}

159−

160− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

161−

162− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

163−

164− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

165−

166− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

167−

168− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

169−

170− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

171−

172− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

173−

174− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

175−

176− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

177−

178− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

179−

180− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

181−

182− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

183−

184− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

185−

186− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Potential"},"finish_reason":null}]}

187−

188− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

189−

190− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

191−

192− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

193−

194− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Using"},"finish_reason":null}]}

195−

196− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ls"},"finish_reason":null}]}

197−

198− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

199−

200− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" List"},"finish_reason":null}]}

201−

202− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Current"},"finish_reason":null}]}

203−

204− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Directory"},"finish_reason":null}]}

205−

206− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Files"},"finish_reason":null}]}

207−

208− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

209−

210− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

211−

212− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

213−

214− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

215−

216− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

217−

218− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

219−

220− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

221−

222− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" colon"},"finish_reason":null}]}

223−

224− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

225−

226− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

227−

228− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

229−

230− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

231−

232− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

233−

234− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

235−

236− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Using"},"finish_reason":null}]}

237−

238− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" ls"},"finish_reason":null}]}

239−

240− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

241−

242− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" List"},"finish_reason":null}]}

243−

244− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Current"},"finish_reason":null}]}

245−

246− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Directory"},"finish_reason":null}]}

247−

248− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Files"},"finish_reason":null}]}

249−

250− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

251−

252− data: {"id":"chatcmpl-7800eed68adf0e90119be770","object":"chat.completion.chunk","created":1778597373,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":21,"completion_tokens":120,"total_tokens":333,"prompt_tokens_details":{"cached_tokens":192}}}

158+ data: {"id":"chatcmpl-5257bdd2dd07faca3a39a451","object":"chat.completion.chunk","created":1778601737,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":33,"completion_tokens":73,"total_tokens":282,"prompt_tokens_details":{"cached_tokens":176}}}

253159

254160 data: [DONE]

255161

@@ -258,15 +164,15 @@ interactions:

258164 - text/event-stream

259165 status: 200 OK

260166 code: 200

261− duration: 458.0435ms

167+ duration: 536.305667ms

262168- id: 1

263169 request:

264170 proto: HTTP/1.1

265171 proto_major: 1

266172 proto_minor: 1

267− content_length: 33637

173+ content_length: 33724

268174 host: ""

269− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

175+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

270176 headers:

271177 Accept:

272178 - application/json

@@ -283,17 +189,41 @@ interactions:

283189 content_length: -1

284190 uncompressed: true

285191 body: |+

286− data: {"id":"chatcmpl-3c3df60659c978a4a68c4826","object":"chat.completion.chunk","created":1778597377,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

193+

194+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_b71d148a2ebc4ca6a2f3a4e2","type":"function","index":0,"function":{"name":"ls","arguments":""}}]},"finish_reason":null}]}

195+

196+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

197+

198+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

199+

200+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"path\": "}}]},"finish_reason":null}]}

201+

202+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

203+

204+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

205+

206+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test/Test"}}]},"finish_reason":null}]}

207+

208+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

209+

210+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

211+

212+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-"}}]},"finish_reason":null}]}

213+

214+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5."}}]},"finish_reason":null}]}

215+

216+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1/ls"}}]},"finish_reason":null}]}

287217

288− data: {"id":"chatcmpl-3c3df60659c978a4a68c4826","object":"chat.completion.chunk","created":1778597377,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_7d06be2f253c411fb9774a21","type":"function","index":0,"function":{"name":"ls","arguments":""}}]},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}]}

289219

290− data: {"id":"chatcmpl-3c3df60659c978a4a68c4826","object":"chat.completion.chunk","created":1778597377,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

291221

292− data: {"id":"chatcmpl-3c3df60659c978a4a68c4826","object":"chat.completion.chunk","created":1778597377,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{}"}}]},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

293223

294− data: {"id":"chatcmpl-3c3df60659c978a4a68c4826","object":"chat.completion.chunk","created":1778597377,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

224+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

295225

296− data: {"id":"chatcmpl-3c3df60659c978a4a68c4826","object":"chat.completion.chunk","created":1778597377,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":702,"completion_tokens":5,"total_tokens":7747,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":4}}}

226+ data: {"id":"chatcmpl-cc9eaa6841f50bdb2bb1138a","object":"chat.completion.chunk","created":1778601741,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7767,"completion_tokens":26,"total_tokens":7793,"completion_tokens_details":{"reasoning_tokens":25}}}

297227

298228 data: [DONE]

299229

@@ -302,15 +232,15 @@ interactions:

302232 - text/event-stream

303233 status: 200 OK

304234 code: 200

305− duration: 5.301269916s

235+ duration: 4.996589333s

306236- id: 2

307237 request:

308238 proto: HTTP/1.1

309239 proto_major: 1

310240 proto_minor: 1

311− content_length: 33926

241+ content_length: 34073

312242 host: ""

313− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_7d06be2f253c411fb9774a21\",\"function\":{\"arguments\":\"{}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/glm-5.1/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_7d06be2f253c411fb9774a21\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

243+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_b71d148a2ebc4ca6a2f3a4e2\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/ls_tool\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/glm-5.1/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_b71d148a2ebc4ca6a2f3a4e2\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

314244 headers:

315245 Accept:

316246 - application/json

@@ -327,27 +257,27 @@ interactions:

327257 content_length: -1

328258 uncompressed: true

329259 body: |+

330− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

331261

332− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Files"},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Files"},"finish_reason":null}]}

333263

334− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" in current"},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

335265

336− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" directory:\n"},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" the current directory"},"finish_reason":null}]}

337267

338− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"- `go"},"finish_reason":null}]}

268+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":\n- `"},"finish_reason":null}]}

339269

340− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}]}

270+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"go.mod`\n"},"finish_reason":null}]}

341271

342− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`\n-"},"finish_reason":null}]}

272+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"- `"},"finish_reason":null}]}

343273

344− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `main"},"finish_reason":null}]}

274+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main.go"},"finish_reason":null}]}

345275

346− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go`"},"finish_reason":null}]}

276+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`"},"finish_reason":null}]}

347277

348− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

278+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

349279

350− data: {"id":"chatcmpl-e09172c463a28df67712f05f","object":"chat.completion.chunk","created":1778597381,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":741,"completion_tokens":17,"total_tokens":7798,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":16}}}

280+ data: {"id":"chatcmpl-f2baa404e30bd65bbb71763f","object":"chat.completion.chunk","created":1778601744,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7827,"completion_tokens":18,"total_tokens":7845,"completion_tokens_details":{"reasoning_tokens":17}}}

351281

352282 data: [DONE]

353283

@@ -356,4 +286,4 @@ interactions:

356286 - text/event-stream

357287 status: 200 OK

358288 code: 200

359− duration: 3.382412042s

289+ duration: 2.545317042s

internal/agent/testdata/TestCoderAgent/glm-5.1/multiedit_tool.yaml+558−538

@@ -25,769 +25,743 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"We"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" produce"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summar"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"izing"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" using"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mult"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mult"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" modify"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" change"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" string"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" World"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"!'"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Use"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mult"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"!'"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" update"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '//"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Greeting"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" above"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" fmt"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".Println"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" should"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Use"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mult"},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" update"},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" U"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" something"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" like"},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")s"},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"..."},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Modify"},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

177177

178− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

178+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

179179

180− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" World"},"finish_reason":null}]}

180+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

181181

182− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

182+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

183183

184− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

184+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Use"},"finish_reason":null}]}

185185

186− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

186+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mult"},"finish_reason":null}]}

187187

188− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

188+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

189189

190− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

190+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

191191

192− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

193193

194− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

194+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" update"},"finish_reason":null}]}

195195

196− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

196+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

197197

198− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

198+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

199199

200− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

200+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

201201

202− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

202+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

203203

204− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

205205

206− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Modify"},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

207207

208− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

209209

210− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" World"},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

211211

212− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

213213

214− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" spaces"},"finish_reason":null}]}

215215

216− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

217217

218− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" punctuation"},"finish_reason":null}]}

219219

220− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

221221

222− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" We'll"},"finish_reason":null}]}

223223

224− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

224+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

225225

226− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

227227

228− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" length"},"finish_reason":null}]}

228+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Use"},"finish_reason":null}]}

229229

230− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"?"},"finish_reason":null}]}

230+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

231231

232− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

232+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

233233

234− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

234+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

235235

236− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" +"},"finish_reason":null}]}

237237

238− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" M"},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

239239

240− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

241241

242− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

243243

244− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")o"},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

245245

246− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

247247

248− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"d"},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

249249

250− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

250+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mult"},"finish_reason":null}]}

251251

252− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"i"},"finish_reason":null}]}

252+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

253253

254− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

254+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

255255

256− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"f"},"finish_reason":null}]}

256+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

257257

258− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

258+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"9"},"finish_reason":null}]}

259259

260− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"y"},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

261261

262− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

263263

264− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"13"},"finish_reason":null}]}

265265

266− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

267267

268− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

268+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

269269

270− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

270+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

271271

272− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"H"},"finish_reason":null}]}

272+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

273273

274− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

274+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

275275

276− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

276+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"14"},"finish_reason":null}]}

277277

278− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"9"},"finish_reason":null}]}

278+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

279279

280− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"l"},"finish_reason":null}]}

280+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

281281

282− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"10"},"finish_reason":null}]}

282+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

283283

284− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"l"},"finish_reason":null}]}

284+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

285285

286− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"11"},"finish_reason":null}]}

286+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

287287

288− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

288+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"16"},"finish_reason":null}]}

289289

290− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"12"},"finish_reason":null}]}

290+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

291291

292− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

292+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

293293

294− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

294+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

295295

296− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

296+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

297297

298− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"13"},"finish_reason":null}]}

298+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

299299

300− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"W"},"finish_reason":null}]}

300+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

301301

302− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"14"},"finish_reason":null}]}

302+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

303303

304− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

304+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" update"},"finish_reason":null}]}

305305

306− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"15"},"finish_reason":null}]}

306+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

307307

308− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"r"},"finish_reason":null}]}

308+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

309309

310− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"16"},"finish_reason":null}]}

310+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

311311

312− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"l"},"finish_reason":null}]}

312+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"23"},"finish_reason":null}]}

313313

314− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

314+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

315315

316− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"d"},"finish_reason":null}]}

316+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

317317

318− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"18"},"finish_reason":null}]}

318+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

319319

320− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

320+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

321321

322− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

322+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

323323

324− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

324+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"24"},"finish_reason":null}]}

325325

326− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"19"},"finish_reason":null}]}

326+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

327327

328− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"t"},"finish_reason":null}]}

328+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

329329

330− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"20"},"finish_reason":null}]}

330+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

331331

332− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

332+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

333333

334− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"21"},"finish_reason":null}]}

334+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

335335

336− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

336+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"32"},"finish_reason":null}]}

337337

338− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

338+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

339339

340− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

340+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

341341

342− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"22"},"finish_reason":null}]}

342+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

343343

344− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"H"},"finish_reason":null}]}

344+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

345345

346− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"23"},"finish_reason":null}]}

346+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

347347

348− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

348+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

349349

350− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"24"},"finish_reason":null}]}

350+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

351351

352− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"l"},"finish_reason":null}]}

352+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

353353

354− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"25"},"finish_reason":null}]}

354+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

355355

356− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"l"},"finish_reason":null}]}

356+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

357357

358− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"26"},"finish_reason":null}]}

358+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

359359

360− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

360+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"36"},"finish_reason":null}]}

361361

362− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"27"},"finish_reason":null}]}

362+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

363363

364− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

364+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

365365

366− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

366+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

367367

368− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

368+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

369369

370− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"28"},"finish_reason":null}]}

370+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

371371

372− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"C"},"finish_reason":null}]}

372+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"37"},"finish_reason":null}]}

373373

374− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"29"},"finish_reason":null}]}

374+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

375375

376− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"r"},"finish_reason":null}]}

376+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

377377

378− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"30"},"finish_reason":null}]}

378+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

379379

380− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"u"},"finish_reason":null}]}

380+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

381381

382− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"31"},"finish_reason":null}]}

382+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

383383

384− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"s"},"finish_reason":null}]}

384+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"40"},"finish_reason":null}]}

385385

386− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"32"},"finish_reason":null}]}

386+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

387387

388− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"h"},"finish_reason":null}]}

388+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

389389

390− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

390+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

391391

392− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

392+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

393393

394− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

394+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

395395

396− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

396+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"41"},"finish_reason":null}]}

397397

398− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"34"},"finish_reason":null}]}

398+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

399399

400− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"a"},"finish_reason":null}]}

400+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

401401

402− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"35"},"finish_reason":null}]}

402+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

403403

404− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"n"},"finish_reason":null}]}

404+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

405405

406− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"36"},"finish_reason":null}]}

406+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

407407

408− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"d"},"finish_reason":null}]}

408+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"48"},"finish_reason":null}]}

409409

410− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"37"},"finish_reason":null}]}

410+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

411411

412− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

412+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

413413

414− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

414+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

415415

416− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

416+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

417417

418− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

418+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

419419

420− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"a"},"finish_reason":null}]}

420+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"49"},"finish_reason":null}]}

421421

422− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"39"},"finish_reason":null}]}

422+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

423423

424− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"d"},"finish_reason":null}]}

424+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

425425

426− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"40"},"finish_reason":null}]}

426+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

427427

428− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"d"},"finish_reason":null}]}

428+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

429429

430− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"41"},"finish_reason":null}]}

430+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

431431

432− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

432+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"51"},"finish_reason":null}]}

433433

434− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

434+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

435435

436− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

436+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

437437

438− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"42"},"finish_reason":null}]}

438+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

439439

440− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"g"},"finish_reason":null}]}

440+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

441441

442− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"43"},"finish_reason":null}]}

442+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

443443

444− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"r"},"finish_reason":null}]}

444+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"52"},"finish_reason":null}]}

445445

446− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"44"},"finish_reason":null}]}

446+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

447447

448− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

448+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

449449

450− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"45"},"finish_reason":null}]}

450+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

451451

452− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

452+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

453453

454− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"46"},"finish_reason":null}]}

454+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

455455

456− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"t"},"finish_reason":null}]}

456+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

457457

458− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"47"},"finish_reason":null}]}

458+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"59"},"finish_reason":null}]}

459459

460− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"i"},"finish_reason":null}]}

460+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

461461

462− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"48"},"finish_reason":null}]}

462+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Too"},"finish_reason":null}]}

463463

464− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"n"},"finish_reason":null}]}

464+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" long"},"finish_reason":null}]}

465465

466− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"49"},"finish_reason":null}]}

466+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

467467

468− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"g"},"finish_reason":null}]}

468+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Need"},"finish_reason":null}]}

469469

470− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

470+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

471471

472− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

472+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

473473

474− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

474+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

475475

476− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

476+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Maybe"},"finish_reason":null}]}

477477

478− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"51"},"finish_reason":null}]}

478+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

479479

480− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"c"},"finish_reason":null}]}

480+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

481481

482− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"52"},"finish_reason":null}]}

482+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Mult"},"finish_reason":null}]}

483483

484− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"o"},"finish_reason":null}]}

484+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

485485

486− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"53"},"finish_reason":null}]}

486+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

487487

488− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"m"},"finish_reason":null}]}

488+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

489489

490− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"54"},"finish_reason":null}]}

490+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" change"},"finish_reason":null}]}

491491

492− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"m"},"finish_reason":null}]}

492+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

493493

494− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"55"},"finish_reason":null}]}

494+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

495495

496− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"e"},"finish_reason":null}]}

496+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

497497

498− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"56"},"finish_reason":null}]}

498+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

499499

500− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"n"},"finish_reason":null}]}

500+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

501501

502− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"57"},"finish_reason":null}]}

502+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

503503

504− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"t"},"finish_reason":null}]}

504+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

505505

506− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"58"},"finish_reason":null}]}

506+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Mult"},"finish_reason":null}]}

507507

508− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

508+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ied"},"finish_reason":null}]}

509509

510− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

510+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"it"},"finish_reason":null}]}

511511

512− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

512+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":("},"finish_reason":null}]}

513513

514− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"58"},"finish_reason":null}]}

514+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"9"},"finish_reason":null}]}

515515

516− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003e"},"finish_reason":null}]}

516+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

517517

518− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

518+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" +"},"finish_reason":null}]}

519519

520− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

520+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

521521

522− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"We"},"finish_reason":null}]}

522+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

523523

524− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

524+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

525525

526− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

526+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"10"},"finish_reason":null}]}

527527

528− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

528+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

529529

530− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

530+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" change"},"finish_reason":null}]}

531531

532− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Perhaps"},"finish_reason":null}]}

532+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

533533

534− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

534+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

535535

536− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Change"},"finish_reason":null}]}

536+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

537537

538− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

538+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"16"},"finish_reason":null}]}

539539

540− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" World"},"finish_reason":null}]}

540+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

541541

542− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

542+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

543543

544− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

544+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

545545

546− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

546+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

547547

548− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

548+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

549549

550− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

550+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

551551

552− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

552+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

553553

554− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

554+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" greeting"},"finish_reason":null}]}

555555

556− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

556+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

557557

558− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

558+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

559559

560− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Change"},"finish_reason":null}]}

560+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

561561

562− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

562+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"25"},"finish_reason":null}]}

563563

564− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

564+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

565565

566− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

566+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

567567

568− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

568+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

569569

570− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

570+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

571571

572− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

572+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

573573

574− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

574+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"26"},"finish_reason":null}]}

575575

576− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

576+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

577577

578− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

578+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

579579

580− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

580+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

581581

582− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

582+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

583583

584− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

584+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

585585

586− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"12"},"finish_reason":null}]}

586+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"29"},"finish_reason":null}]}

587587

588− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

588+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

589589

590− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

590+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

591591

592− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"13"},"finish_reason":null}]}

592+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

593593

594− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

594+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

595595

596− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" World"},"finish_reason":null}]}

596+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

597597

598− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

598+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"30"},"finish_reason":null}]}

599599

600− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

600+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

601601

602− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

602+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

603603

604− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"18"},"finish_reason":null}]}

604+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

605605

606− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

606+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

607607

608− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

608+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

609609

610− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"19"},"finish_reason":null}]}

610+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

611611

612− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

612+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

613613

614− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

614+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

615615

616− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

616+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

617617

618− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

618+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

619619

620− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

620+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

621621

622− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"21"},"finish_reason":null}]}

622+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"34"},"finish_reason":null}]}

623623

624− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

624+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

625625

626− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

626+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

627627

628− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"22"},"finish_reason":null}]}

628+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

629629

630− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

630+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

631631

632− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

632+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

633633

634− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

634+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"41"},"finish_reason":null}]}

635635

636− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

636+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

637637

638− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

638+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

639639

640− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"27"},"finish_reason":null}]}

640+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

641641

642− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

642+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"41"},"finish_reason":null}]}

643643

644− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

644+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

645645

646− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"28"},"finish_reason":null}]}

646+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" within"},"finish_reason":null}]}

647647

648− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

648+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" limit"},"finish_reason":null}]}

649649

650− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

650+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

651651

652− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

652+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Return"},"finish_reason":null}]}

653653

654− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

654+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

655655

656− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

656+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

657657

658− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

658+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Mult"},"finish_reason":null}]}

659659

660− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

660+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"ied"},"finish_reason":null}]}

661661

662− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

662+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"it"},"finish_reason":null}]}

663663

664− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"34"},"finish_reason":null}]}

664+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}]}

665665

666− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

666+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" change"},"finish_reason":null}]}

667667

668− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

668+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" greeting"},"finish_reason":null}]}

669669

670− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

670+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" and"},"finish_reason":null}]}

671671

672− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

672+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" add"},"finish_reason":null}]}

673673

674− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

674+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" comment"},"finish_reason":null}]}

675675

676− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"37"},"finish_reason":null}]}

676+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

677677

678− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

678+ data: {"id":"chatcmpl-1681c841a760d422b6d9bbf0","object":"chat.completion.chunk","created":1778601745,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":55,"completion_tokens":333,"total_tokens":564,"prompt_tokens_details":{"cached_tokens":176}}}

679679

680− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

681−

682− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

683−

684− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

685−

686− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

687−

688− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

689−

690− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

691−

692− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

693−

694− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"41"},"finish_reason":null}]}

695−

696− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

697−

698− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

699−

700− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"42"},"finish_reason":null}]}

701−

702− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

703−

704− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

705−

706− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

707−

708− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

709−

710− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

711−

712− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"49"},"finish_reason":null}]}

713−

714− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

715−

716− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

717−

718− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

719−

720− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"49"},"finish_reason":null}]}

721−

722− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

723−

724− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

725−

726− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

727−

728− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

729−

730− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

731−

732− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

733−

734− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Change"},"finish_reason":null}]}

735−

736− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

680+ data: [DONE]

737681

738− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" World"},"finish_reason":null}]}

682+ headers:

683+ Content-Type:

684+ - text/event-stream

685+ status: 200 OK

686+ code: 200

687+ duration: 600.672042ms

688+- id: 1

689+ request:

690+ proto: HTTP/1.1

691+ proto_major: 1

692+ proto_minor: 1

693+ content_length: 33810

694+ host: ""

695+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

696+ headers:

697+ Accept:

698+ - application/json

699+ Content-Type:

700+ - application/json

701+ User-Agent:

702+ - Charm-Crush/devel (https://charm.land/crush)

703+ url: https://hyper.charm.land/v1/chat/completions

704+ method: POST

705+ response:

706+ proto: HTTP/2.0

707+ proto_major: 2

708+ proto_minor: 0

709+ content_length: -1

710+ uncompressed: true

711+ body: |+

712+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

739713

740− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

714+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Let"},"finish_reason":null}]}

741715

742− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

716+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me"},"finish_reason":null}]}

743717

744− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

718+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" first read"},"finish_reason":null}]}

745719

746− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

720+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the main.go"},"finish_reason":null}]}

747721

748− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" add"},"finish_reason":null}]}

722+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" file to see"},"finish_reason":null}]}

749723

750− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" comment"},"finish_reason":null}]}

724+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" its current"},"finish_reason":null}]}

751725

752− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\n\n"},"finish_reason":null}]}

726+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" contents."},"finish_reason":null}]}

753727

754− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"No"},"finish_reason":null}]}

728+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_813fe8edde9041c18e85fd3c","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

755729

756− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

730+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

757731

758− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

732+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

759733

760− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" colon"},"finish_reason":null}]}

734+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

761735

762− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

736+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

763737

764− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" One"},"finish_reason":null}]}

738+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

765739

766− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

740+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

767741

768− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

742+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

769743

770− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Change"},"finish_reason":null}]}

744+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

771745

772− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Hello"},"finish_reason":null}]}

746+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

773747

774− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" World"},"finish_reason":null}]}

748+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5.1"}}]},"finish_reason":null}]}

775749

776− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

750+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/mult"}}]},"finish_reason":null}]}

777751

778− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Hello"},"finish_reason":null}]}

752+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}]}

779753

780− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Crush"},"finish_reason":null}]}

754+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}]}

781755

782− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" and"},"finish_reason":null}]}

756+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool/main.go"}}]},"finish_reason":null}]}

783757

784− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" add"},"finish_reason":null}]}

758+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

785759

786− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" comment"},"finish_reason":null}]}

760+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

787761

788− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

762+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

789763

790− data: {"id":"chatcmpl-c33fa86d503c142b247f086f","object":"chat.completion.chunk","created":1778597382,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":43,"completion_tokens":389,"total_tokens":624,"prompt_tokens_details":{"cached_tokens":192}}}

764+ data: {"id":"chatcmpl-d7f66a4943ebca0c1fa332dd","object":"chat.completion.chunk","created":1778601748,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7792,"completion_tokens":45,"total_tokens":7837,"completion_tokens_details":{"reasoning_tokens":14}}}

791765

792766 data: [DONE]

793767

@@ -796,15 +770,15 @@ interactions:

796770 - text/event-stream

797771 status: 200 OK

798772 code: 200

799− duration: 525.624875ms

800−- id: 1

773+ duration: 3.563327417s

774+- id: 2

801775 request:

802776 proto: HTTP/1.1

803777 proto_major: 1

804778 proto_minor: 1

805− content_length: 33723

779+ content_length: 34341

806780 host: ""

807− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

781+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_813fe8edde9041c18e85fd3c\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me first read the main.go file to see its current contents.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_813fe8edde9041c18e85fd3c\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

808782 headers:

809783 Accept:

810784 - application/json

@@ -821,65 +795,99 @@ interactions:

821795 content_length: -1

822796 uncompressed: true

823797 body: |+

824− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

798+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

799+

800+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_c18af77e70014616b852dd7f","type":"function","index":0,"function":{"name":"multiedit","arguments":""}}]},"finish_reason":null}]}

801+

802+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

803+

804+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

805+

806+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"edits\": "}}]},"finish_reason":null}]}

807+

808+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"[{\""}}]},"finish_reason":null}]}

809+

810+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new"}}]},"finish_reason":null}]}

811+

812+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string\": \""}}]},"finish_reason":null}]}

813+

814+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}]}

815+

816+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", Crush"}}]},"finish_reason":null}]}

817+

818+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\", \""}}]},"finish_reason":null}]}

819+

820+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old_string\":"}}]},"finish_reason":null}]}

825821

826− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Let"},"finish_reason":null}]}

822+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \"Hello"}}]},"finish_reason":null}]}

827823

828− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me"},"finish_reason":null}]}

824+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", World"}}]},"finish_reason":null}]}

829825

830− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" first read"},"finish_reason":null}]}

826+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\"}, {\""}}]},"finish_reason":null}]}

831827

832− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the main.go"},"finish_reason":null}]}

828+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new_string\":"}}]},"finish_reason":null}]}

833829

834− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

830+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \"\\t"}}]},"finish_reason":null}]}

835831

836− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

832+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"// G"}}]},"finish_reason":null}]}

837833

838− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" see its"},"finish_reason":null}]}

834+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"reeting\\n\\t"}}]},"finish_reason":null}]}

839835

840− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" current contents."},"finish_reason":null}]}

836+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt.Println"}}]},"finish_reason":null}]}

841837

842− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_283007539bbb437197ef527d","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

838+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}]}

843839

844− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

840+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello, Crush"}}]},"finish_reason":null}]}

845841

846− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

842+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")"}}]},"finish_reason":null}]}

847843

848− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": \"/tmp"}}]},"finish_reason":null}]}

844+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", \""}}]},"finish_reason":null}]}

849845

850− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

846+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old_string\":"}}]},"finish_reason":null}]}

851847

852− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

848+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \"\\t"}}]},"finish_reason":null}]}

853849

854− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

850+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt.Println(\\\""}}]},"finish_reason":null}]}

855851

856− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

852+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello, World"}}]},"finish_reason":null}]}

857853

858− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

854+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")"}}]},"finish_reason":null}]}

859855

860− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

856+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}]"}}]},"finish_reason":null}]}

861857

862− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm"}}]},"finish_reason":null}]}

858+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

863859

864− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

860+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

865861

866− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1/m"}}]},"finish_reason":null}]}

862+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

867863

868− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ultied"}}]},"finish_reason":null}]}

864+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

869865

870− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}]}

866+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test/Test"}}]},"finish_reason":null}]}

871867

872− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}]}

868+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

873869

874− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go"}}]},"finish_reason":null}]}

870+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

875871

876− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

872+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

877873

878− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

874+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

879875

880− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

876+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1"}}]},"finish_reason":null}]}

881877

882− data: {"id":"chatcmpl-78dc0a02c287e4423ccf7646","object":"chat.completion.chunk","created":1778597385,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7767,"completion_tokens":45,"total_tokens":7812,"completion_tokens_details":{"reasoning_tokens":14}}}

878+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/mult"}}]},"finish_reason":null}]}

879+

880+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"iedit"}}]},"finish_reason":null}]}

881+

882+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool/main"}}]},"finish_reason":null}]}

883+

884+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go\""}}]},"finish_reason":null}]}

885+

886+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

887+

888+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

889+

890+ data: {"id":"chatcmpl-a69a664769bb94eb14ee9e5d","object":"chat.completion.chunk","created":1778601751,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7879,"completion_tokens":95,"total_tokens":7974,"completion_tokens_details":{"reasoning_tokens":94}}}

883891

884892 data: [DONE]

885893

@@ -888,15 +896,15 @@ interactions:

888896 - text/event-stream

889897 status: 200 OK

890898 code: 200

891− duration: 4.180391958s

892−- id: 2

899+ duration: 2.000582291s

900+- id: 3

893901 request:

894902 proto: HTTP/1.1

895903 proto_major: 1

896904 proto_minor: 1

897− content_length: 34254

905+ content_length: 35017

898906 host: ""

899− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_283007539bbb437197ef527d\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me first read the main.go file to see its current contents.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_283007539bbb437197ef527d\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

907+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_813fe8edde9041c18e85fd3c\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me first read the main.go file to see its current contents.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_813fe8edde9041c18e85fd3c\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_c18af77e70014616b852dd7f\",\"function\":{\"arguments\":\"{\\\"edits\\\": [{\\\"new_string\\\": \\\"Hello, Crush!\\\", \\\"old_string\\\": \\\"Hello, World!\\\"}, {\\\"new_string\\\": \\\"\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\"}], \\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nApplied 1 of 2 edits to file: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go (1 edit(s) failed)\\n</result>\\n\",\"tool_call_id\":\"call_c18af77e70014616b852dd7f\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

900908 headers:

901909 Accept:

902910 - application/json

@@ -913,83 +921,79 @@ interactions:

913921 content_length: -1

914922 uncompressed: true

915923 body: |+

916− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

917−

918− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_438a3edf38b3453e9b16eb16","type":"function","index":0,"function":{"name":"multiedit","arguments":""}}]},"finish_reason":null}]}

924+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

919925

920− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

926+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"The"},"finish_reason":null}]}

921927

922− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

928+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" first edit"},"finish_reason":null}]}

923929

924− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"edits\": "}}]},"finish_reason":null}]}

930+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" changed"},"finish_reason":null}]}

925931

926− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"[{\"new_string"}}]},"finish_reason":null}]}

932+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" the"},"finish_reason":null}]}

927933

928− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\": \"\\"}}]},"finish_reason":null}]}

934+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" string, which"},"finish_reason":null}]}

929935

930− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"t//"}}]},"finish_reason":null}]}

936+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" made"},"finish_reason":null}]}

931937

932− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" G"}}]},"finish_reason":null}]}

938+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" the second"},"finish_reason":null}]}

933939

934− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"reeting"}}]},"finish_reason":null}]}

940+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" edit"},"finish_reason":null}]}

935941

936− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n\\t"}}]},"finish_reason":null}]}

942+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"'s old"},"finish_reason":null}]}

937943

938− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt.Println"}}]},"finish_reason":null}]}

944+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_string"},"finish_reason":null}]}

939945

940− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\"Hello"}}]},"finish_reason":null}]}

946+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" not"},"finish_reason":null}]}

941947

942− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", Crush"}}]},"finish_reason":null}]}

948+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" match. Let"},"finish_reason":null}]}

943949

944− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")"}}]},"finish_reason":null}]}

950+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" me re"},"finish_reason":null}]}

945951

946− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", \"old"}}]},"finish_reason":null}]}

952+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-read and"},"finish_reason":null}]}

947953

948− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string\": \"\\"}}]},"finish_reason":null}]}

954+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" fix"},"finish_reason":null}]}

949955

950− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"t"}}]},"finish_reason":null}]}

956+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"."},"finish_reason":null}]}

951957

952− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt.Println(\\\"Hello"}}]},"finish_reason":null}]}

958+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_72491b955eb045b08f557741","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

953959

954− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", World"}}]},"finish_reason":null}]}

960+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

955961

956− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")"}}]},"finish_reason":null}]}

962+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

957963

958− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}]"}}]},"finish_reason":null}]}

964+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

959965

960− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

966+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

961967

962− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

968+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

963969

964− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp"}}]},"finish_reason":null}]}

970+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

965971

966− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

972+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

967973

968− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

974+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

969975

970− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test/Test"}}]},"finish_reason":null}]}

976+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-"}}]},"finish_reason":null}]}

971977

972− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"CoderAgent"}}]},"finish_reason":null}]}

978+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5."}}]},"finish_reason":null}]}

973979

974− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/g"}}]},"finish_reason":null}]}

980+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1/m"}}]},"finish_reason":null}]}

975981

976− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-"}}]},"finish_reason":null}]}

982+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}]}

977983

978− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5.1/m"}}]},"finish_reason":null}]}

984+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}]}

979985

980− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ultied"}}]},"finish_reason":null}]}

986+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it_tool"}}]},"finish_reason":null}]}

981987

982− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}]}

988+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go"}}]},"finish_reason":null}]}

983989

984− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool/main"}}]},"finish_reason":null}]}

990+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

985991

986− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go\""}}]},"finish_reason":null}]}

992+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

987993

988− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

994+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

989995

990− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

991−

992− data: {"id":"chatcmpl-2474e58b91064f2059f283db","object":"chat.completion.chunk","created":1778597390,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":814,"completion_tokens":76,"total_tokens":7930,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":75}}}

996+ data: {"id":"chatcmpl-2edb4a948db2acecbbcac1b7","object":"chat.completion.chunk","created":1778601756,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":8022,"completion_tokens":56,"total_tokens":8078,"completion_tokens_details":{"reasoning_tokens":55}}}

993997

994998 data: [DONE]

995999

@@ -998,15 +1002,15 @@ interactions:

9981002 - text/event-stream

9991003 status: 200 OK

10001004 code: 200

1001− duration: 4.191709167s

1002−- id: 3

1005+ duration: 3.717709083s

1006+- id: 4

10031007 request:

10041008 proto: HTTP/1.1

10051009 proto_major: 1

10061010 proto_minor: 1

1007− content_length: 34834

1011+ content_length: 35607

10081012 host: ""

1009− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_283007539bbb437197ef527d\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me first read the main.go file to see its current contents.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_283007539bbb437197ef527d\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_438a3edf38b3453e9b16eb16\",\"function\":{\"arguments\":\"{\\\"edits\\\": [{\\\"new_string\\\": \\\"\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\"}], \\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\n</result>\\n\",\"tool_call_id\":\"call_438a3edf38b3453e9b16eb16\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

1013+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_813fe8edde9041c18e85fd3c\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me first read the main.go file to see its current contents.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_813fe8edde9041c18e85fd3c\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_c18af77e70014616b852dd7f\",\"function\":{\"arguments\":\"{\\\"edits\\\": [{\\\"new_string\\\": \\\"Hello, Crush!\\\", \\\"old_string\\\": \\\"Hello, World!\\\"}, {\\\"new_string\\\": \\\"\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\"}], \\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nApplied 1 of 2 edits to file: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go (1 edit(s) failed)\\n</result>\\n\",\"tool_call_id\":\"call_c18af77e70014616b852dd7f\",\"role\":\"tool\"},{\"content\":\"The first edit changed the string, which made the second edit's old_string not match. Let me re-read and fix.\",\"tool_calls\":[{\"id\":\"call_72491b955eb045b08f557741\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, Crush!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_72491b955eb045b08f557741\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

10101014 headers:

10111015 Accept:

10121016 - application/json

@@ -1023,45 +1027,63 @@ interactions:

10231027 content_length: -1

10241028 uncompressed: true

10251029 body: |+

1026− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

1030+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

10271031

1028− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_a17a720072ed49faa4627ce8","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

1032+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_e5a1488c26a84022bb580a81","type":"function","index":0,"function":{"name":"edit","arguments":""}}]},"finish_reason":null}]}

10291033

1030− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

1034+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

10311035

1032− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

1036+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

10331037

1034− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

1038+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": \"/tmp"}}]},"finish_reason":null}]}

10351039

1036− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

1040+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

10371041

1038− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

1042+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

10391043

1040− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

1044+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

10411045

1042− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

1046+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

10431047

1044− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

1048+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

10451049

1046− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-5"}}]},"finish_reason":null}]}

1050+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-"}}]},"finish_reason":null}]}

10471051

1048− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1/m"}}]},"finish_reason":null}]}

1052+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5.1/m"}}]},"finish_reason":null}]}

10491053

1050− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ultied"}}]},"finish_reason":null}]}

1054+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ultiedit_tool"}}]},"finish_reason":null}]}

10511055

1052− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}]}

1056+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go\""}}]},"finish_reason":null}]}

10531057

1054− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}]}

1058+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

10551059

1056− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go"}}]},"finish_reason":null}]}

1060+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"new_string\": "}}]},"finish_reason":null}]}

10571061

1058− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

1062+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"\\t// G"}}]},"finish_reason":null}]}

10591063

1060− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

1064+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"reeting\\n\\tfmt"}}]},"finish_reason":null}]}

10611065

1062− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

1066+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println(\\\"Hello"}}]},"finish_reason":null}]}

10631067

1064− data: {"id":"chatcmpl-1c84d2444240539e0bffc35c","object":"chat.completion.chunk","created":1778597395,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":225,"completion_tokens":31,"total_tokens":8000,"prompt_tokens_details":{"cached_tokens":7744},"completion_tokens_details":{"reasoning_tokens":30}}}

1068+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", Crush"}}]},"finish_reason":null}]}

1069+

1070+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")\", "}}]},"finish_reason":null}]}

1071+

1072+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"old_string\": "}}]},"finish_reason":null}]}

1073+

1074+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"\\tfmt"}}]},"finish_reason":null}]}

1075+

1076+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println(\\\"Hello,"}}]},"finish_reason":null}]}

1077+

1078+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Crush!\\\")"}}]},"finish_reason":null}]}

1079+

1080+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

1081+

1082+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

1083+

1084+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

1085+

1086+ data: {"id":"chatcmpl-d163029b902928d9f8c7e6da","object":"chat.completion.chunk","created":1778601763,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":8134,"completion_tokens":62,"total_tokens":8196,"completion_tokens_details":{"reasoning_tokens":61}}}

10651087

10661088 data: [DONE]

10671089

@@ -1070,15 +1092,15 @@ interactions:

10701092 - text/event-stream

10711093 status: 200 OK

10721094 code: 200

1073− duration: 3.261922667s

1074−- id: 4

1095+ duration: 4.809219875s

1096+- id: 5

10751097 request:

10761098 proto: HTTP/1.1

10771099 proto_major: 1

10781100 proto_minor: 1

1079− content_length: 35324

1101+ content_length: 36168

10801102 host: ""

1081− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_283007539bbb437197ef527d\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me first read the main.go file to see its current contents.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_283007539bbb437197ef527d\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_438a3edf38b3453e9b16eb16\",\"function\":{\"arguments\":\"{\\\"edits\\\": [{\\\"new_string\\\": \\\"\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\"}], \\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\n</result>\\n\",\"tool_call_id\":\"call_438a3edf38b3453e9b16eb16\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_a17a720072ed49faa4627ce8\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\t// Greeting\\n 7|\\tfmt.Println(\\\"Hello, Crush!\\\")\\n 8|}\\n</file>\\n\",\"tool_call_id\":\"call_a17a720072ed49faa4627ce8\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

1103+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_813fe8edde9041c18e85fd3c\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me first read the main.go file to see its current contents.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_813fe8edde9041c18e85fd3c\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_c18af77e70014616b852dd7f\",\"function\":{\"arguments\":\"{\\\"edits\\\": [{\\\"new_string\\\": \\\"Hello, Crush!\\\", \\\"old_string\\\": \\\"Hello, World!\\\"}, {\\\"new_string\\\": \\\"\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\"}], \\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nApplied 1 of 2 edits to file: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go (1 edit(s) failed)\\n</result>\\n\",\"tool_call_id\":\"call_c18af77e70014616b852dd7f\",\"role\":\"tool\"},{\"content\":\"The first edit changed the string, which made the second edit's old_string not match. Let me re-read and fix.\",\"tool_calls\":[{\"id\":\"call_72491b955eb045b08f557741\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, Crush!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_72491b955eb045b08f557741\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_e5a1488c26a84022bb580a81\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\\", \\\"new_string\\\": \\\"\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/glm-5.1/multiedit_tool/main.go\\n</result>\\n\",\"tool_call_id\":\"call_e5a1488c26a84022bb580a81\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

10821104 headers:

10831105 Accept:

10841106 - application/json

@@ -1095,43 +1117,41 @@ interactions:

10951117 content_length: -1

10961118 uncompressed: true

10971119 body: |+

1098− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

1099−

1100− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}]}

1120+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

11011121

1102− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" —"},"finish_reason":null}]}

1122+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}]}

11031123

1104− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `'"},"finish_reason":null}]}

1124+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" —"},"finish_reason":null}]}

11051125

1106− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}

1126+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" both changes"},"finish_reason":null}]}

11071127

1108− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":", World"},"finish_reason":null}]}

1128+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" applied"},"finish_reason":null}]}

11091129

1110− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!'` changed"},"finish_reason":null}]}

1130+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":": `\""},"finish_reason":null}]}

11111131

1112− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" to `'"},"finish_reason":null}]}

1132+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Hello, World"},"finish_reason":null}]}

11131133

1114− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Hello, Crush"},"finish_reason":null}]}

1134+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!\"` →"},"finish_reason":null}]}

11151135

1116− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!'` and `"},"finish_reason":null}]}

1136+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `\"Hello, Crush"},"finish_reason":null}]}

11171137

1118− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"// G"},"finish_reason":null}]}

1138+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!\"` and `"},"finish_reason":null}]}

11191139

1120− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"reeting"},"finish_reason":null}]}

1140+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"// G"},"finish_reason":null}]}

11211141

1122− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` comment added"},"finish_reason":null}]}

1142+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"reeting` comment"},"finish_reason":null}]}

11231143

1124− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" above"},"finish_reason":null}]}

1144+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" added"},"finish_reason":null}]}

11251145

1126− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

1146+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" above the `"},"finish_reason":null}]}

11271147

1128− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"fmt.Println"},"finish_reason":null}]}

1148+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"fmt.Println` line"},"finish_reason":null}]}

11291149

1130− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`."},"finish_reason":null}]}

1150+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"."},"finish_reason":null}]}

11311151

1132− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

1152+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

11331153

1134− data: {"id":"chatcmpl-b0eeafb8638d7348f03ba44f","object":"chat.completion.chunk","created":1778597397,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":8065,"completion_tokens":31,"total_tokens":8096,"completion_tokens_details":{"reasoning_tokens":30}}}

1154+ data: {"id":"chatcmpl-2b5097e0950d1e7e3132f1fd","object":"chat.completion.chunk","created":1778601766,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":8233,"completion_tokens":37,"total_tokens":8270,"completion_tokens_details":{"reasoning_tokens":36}}}

11351155

11361156 data: [DONE]

11371157

@@ -1140,4 +1160,4 @@ interactions:

11401160 - text/event-stream

11411161 status: 200 OK

11421162 code: 200

1143− duration: 1.715269625s

1163+ duration: 2.961048333s

internal/agent/testdata/TestCoderAgent/glm-5.1/parallel_tool_calls.yaml+337−177

@@ -25,249 +25,371 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" says"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" says"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" all"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" all"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" use"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ls"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" use"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ..."},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ls"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" run"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" both"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" list"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" tool"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" calls"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" current"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" parallel"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"\n\n"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" it"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"We"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" is"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" very"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" important"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" you"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" run"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" both"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" tool"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" calls"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" parallel"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summar"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"izing"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Language"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" English"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Parallel"},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Something"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ls"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" like"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" listing"},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Parallel"},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Glob"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" maybe"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Search"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Parallel"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Directory"},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Listing"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

177177

178− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

178+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

179179

180− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

180+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

181181

182− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" listing"},"finish_reason":null}]}

182+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Parallel"},"finish_reason":null}]}

183183

184− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

184+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Glob"},"finish_reason":null}]}

185185

186− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

186+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

187187

188− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

188+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

189189

190− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

190+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Search"},"finish_reason":null}]}

191191

192− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

193193

194− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

194+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Directory"},"finish_reason":null}]}

195195

196− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

196+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Listing"},"finish_reason":null}]}

197197

198− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

198+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

199199

200− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Parallel"},"finish_reason":null}]}

200+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ="},"finish_reason":null}]}

201201

202− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

202+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" let's"},"finish_reason":null}]}

203203

204− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

205205

206− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

207207

208− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Parallel"},"finish_reason":null}]}

209209

210− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" files"},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

211211

212− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

213213

214− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" directory"},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

215215

216− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" listing"},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" +"},"finish_reason":null}]}

217217

218− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

219219

220− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

221221

222− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

223223

224− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"48"},"finish_reason":null}]}

224+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"9"},"finish_reason":null}]}

225225

226− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

227227

228− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

228+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Glob"},"finish_reason":null}]}

229229

230− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"including"},"finish_reason":null}]}

230+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

231231

232− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" spaces"},"finish_reason":null}]}

232+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

233233

234− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

234+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

235235

236− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"13"},"finish_reason":null}]}

237237

238− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

239239

240− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Return"},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

241241

242− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" just"},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

243243

244− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"14"},"finish_reason":null}]}

245245

246− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

247247

248− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ."},"finish_reason":null}]}

249249

250− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Parallel"},"finish_reason":null}]}

250+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"go"},"finish_reason":null}]}

251251

252− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" glob"},"finish_reason":null}]}

252+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

253253

254− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" for"},"finish_reason":null}]}

254+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

255255

256− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" ."},"finish_reason":null}]}

256+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

257257

258− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}]}

258+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

259259

260− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" files"},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

261261

262− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" and"},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

263263

264− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" directory"},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

265265

266− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" listing"},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"18"},"finish_reason":null}]}

267267

268− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

268+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

269269

270− data: {"id":"chatcmpl-3572456ad6c6fd07d5c164cc","object":"chat.completion.chunk","created":1778597427,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":40,"completion_tokens":129,"total_tokens":361,"prompt_tokens_details":{"cached_tokens":192}}}

270+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Search"},"finish_reason":null}]}

271+

272+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

273+

274+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

275+

276+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

277+

278+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"24"},"finish_reason":null}]}

279+

280+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

281+

282+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

283+

284+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

285+

286+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"25"},"finish_reason":null}]}

287+

288+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

289+

290+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

291+

292+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

293+

294+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

295+

296+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

297+

298+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"28"},"finish_reason":null}]}

299+

300+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

301+

302+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

303+

304+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

305+

306+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"29"},"finish_reason":null}]}

307+

308+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

309+

310+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Directory"},"finish_reason":null}]}

311+

312+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

313+

314+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"9"},"finish_reason":null}]}

315+

316+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

317+

318+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

319+

320+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

321+

322+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

323+

324+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

325+

326+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"39"},"finish_reason":null}]}

327+

328+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

329+

330+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Listing"},"finish_reason":null}]}

331+

332+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

333+

334+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

335+

336+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

337+

338+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"46"},"finish_reason":null}]}

339+

340+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

341+

342+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

343+

344+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

345+

346+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"46"},"finish_reason":null}]}

347+

348+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

349+

350+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

351+

352+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

353+

354+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

355+

356+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"No"},"finish_reason":null}]}

357+

358+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

359+

360+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

361+

362+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" One"},"finish_reason":null}]}

363+

364+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

365+

366+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

367+

368+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Return"},"finish_reason":null}]}

369+

370+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

371+

372+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

373+

374+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Parallel"},"finish_reason":null}]}

375+

376+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Glob"},"finish_reason":null}]}

377+

378+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" ."},"finish_reason":null}]}

379+

380+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}]}

381+

382+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Search"},"finish_reason":null}]}

383+

384+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" and"},"finish_reason":null}]}

385+

386+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Directory"},"finish_reason":null}]}

387+

388+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Listing"},"finish_reason":null}]}

389+

390+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

391+

392+ data: {"id":"chatcmpl-cb928907f856822f9bb954b1","object":"chat.completion.chunk","created":1778601795,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":40,"completion_tokens":190,"total_tokens":422,"prompt_tokens_details":{"cached_tokens":192}}}

271393

272394 data: [DONE]

273395

@@ -276,15 +398,15 @@ interactions:

276398 - text/event-stream

277399 status: 200 OK

278400 code: 200

279− duration: 473.337ms

401+ duration: 477.99625ms

280402- id: 1

281403 request:

282404 proto: HTTP/1.1

283405 proto_major: 1

284406 proto_minor: 1

285− content_length: 33734

407+ content_length: 33821

286408 host: ""

287− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

409+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

288410 headers:

289411 Accept:

290412 - application/json

@@ -301,59 +423,101 @@ interactions:

301423 content_length: -1

302424 uncompressed: true

303425 body: |+

304− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

426+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

427+

428+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

429+

430+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

305431

306− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

432+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me"},"finish_reason":null}]}

307433

308− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

434+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to run"},"finish_reason":null}]}

309435

310− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me"},"finish_reason":null}]}

436+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" two independent"},"finish_reason":null}]}

311437

312− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to run both"},"finish_reason":null}]}

438+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" tool"},"finish_reason":null}]}

313439

314− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" tools"},"finish_reason":null}]}

440+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" calls"},"finish_reason":null}]}

315441

316− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" in parallel -"},"finish_reason":null}]}

442+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" in parallel"},"finish_reason":null}]}

317443

318− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" glob"},"finish_reason":null}]}

444+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":": glob"},"finish_reason":null}]}

319445

320− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

446+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

321447

322− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" find ."},"finish_reason":null}]}

448+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" find ."},"finish_reason":null}]}

323449

324− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"go files"},"finish_reason":null}]}

450+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"go files"},"finish_reason":null}]}

325451

326− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" and ls"},"finish_reason":null}]}

452+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" and ls"},"finish_reason":null}]}

327453

328− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to list"},"finish_reason":null}]}

454+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to list"},"finish_reason":null}]}

329455

330− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the current directory"},"finish_reason":null}]}

456+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the current directory"},"finish_reason":null}]}

331457

332− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

458+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

333459

334− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_db9a486cd35645f3b90ee531","type":"function","index":0,"function":{"name":"glob","arguments":""}}]},"finish_reason":null}]}

460+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_0f6894f9f61447448d4c4684","type":"function","index":0,"function":{"name":"glob","arguments":""}}]},"finish_reason":null}]}

335461

336− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

462+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

337463

338− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

464+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

339465

340− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

466+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

341467

342− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"**/*."}}]},"finish_reason":null}]}

468+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"**/*.go"}}]},"finish_reason":null}]}

343469

344− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go\""}}]},"finish_reason":null}]}

470+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

345471

346− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

472+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"path\": \"/tmp"}}]},"finish_reason":null}]}

347473

348− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_76a6b00364c747e69ca63e2e","type":"function","index":1,"function":{"name":"ls","arguments":""}}]},"finish_reason":null}]}

474+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

349475

350− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":""}}]},"finish_reason":null}]}

476+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

351477

352− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"{}"}}]},"finish_reason":null}]}

478+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

353479

354− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

480+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

355481

356− data: {"id":"chatcmpl-b7b4b066f9d904b86ddbbd70","object":"chat.completion.chunk","created":1778597431,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":722,"completion_tokens":41,"total_tokens":7803,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":25}}}

482+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

483+

484+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

485+

486+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

487+

488+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

489+

490+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1/"}}]},"finish_reason":null}]}

491+

492+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"parallel_tool"}}]},"finish_reason":null}]}

493+

494+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_calls"}}]},"finish_reason":null}]}

495+

496+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

497+

498+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

499+

500+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_8f3552dac9144331b36f56d6","type":"function","index":1,"function":{"name":"ls","arguments":""}}]},"finish_reason":null}]}

501+

502+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":""}}]},"finish_reason":null}]}

503+

504+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"{\"path\": "}}]},"finish_reason":null}]}

505+

506+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"\"/tmp/crush-test"}}]},"finish_reason":null}]}

507+

508+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"/TestCoderAgent/g"}}]},"finish_reason":null}]}

509+

510+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"lm-5."}}]},"finish_reason":null}]}

511+

512+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"1/parallel_tool"}}]},"finish_reason":null}]}

513+

514+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"_calls\""}}]},"finish_reason":null}]}

515+

516+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"}"}}]},"finish_reason":null}]}

517+

518+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

519+

520+ data: {"id":"chatcmpl-48fbb55262ca288daf4e9c41","object":"chat.completion.chunk","created":1778601801,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7787,"completion_tokens":87,"total_tokens":7874,"completion_tokens_details":{"reasoning_tokens":27}}}

357521

358522 data: [DONE]

359523

@@ -362,15 +526,15 @@ interactions:

362526 - text/event-stream

363527 status: 200 OK

364528 code: 200

365− duration: 3.781553792s

529+ duration: 6.487687083s

366530- id: 2

367531 request:

368532 proto: HTTP/1.1

369533 proto_major: 1

370534 proto_minor: 1

371− content_length: 34435

535+ content_length: 34683

372536 host: ""

373− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_db9a486cd35645f3b90ee531\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"call_76a6b00364c747e69ca63e2e\",\"function\":{\"arguments\":\"{}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to run both tools in parallel - glob to find .go files and ls to list the current directory.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls/main.go\",\"tool_call_id\":\"call_db9a486cd35645f3b90ee531\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_76a6b00364c747e69ca63e2e\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

537+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_0f6894f9f61447448d4c4684\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/*.go\\\", \\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"call_8f3552dac9144331b36f56d6\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to run two independent tool calls in parallel: glob to find .go files and ls to list the current directory.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls/main.go\",\"tool_call_id\":\"call_0f6894f9f61447448d4c4684\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/glm-5.1/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_8f3552dac9144331b36f56d6\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

374538 headers:

375539 Accept:

376540 - application/json

@@ -387,43 +551,39 @@ interactions:

387551 content_length: -1

388552 uncompressed: true

389553 body: |+

390− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

391−

392− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"**"},"finish_reason":null}]}

393−

394− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"G"},"finish_reason":null}]}

554+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

395555

396− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"lob results"},"finish_reason":null}]}

556+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}]}

397557

398− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"**: Found"},"finish_reason":null}]}

558+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" **G"},"finish_reason":null}]}

399559

400− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 1 `."},"finish_reason":null}]}

560+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"lob**:"},"finish_reason":null}]}

401561

402− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"go`"},"finish_reason":null}]}

562+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" One"},"finish_reason":null}]}

403563

404− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" file: `"},"finish_reason":null}]}

564+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `."},"finish_reason":null}]}

405565

406− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main.go"},"finish_reason":null}]}

566+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"go`"},"finish_reason":null}]}

407567

408− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`\n\n**"},"finish_reason":null}]}

568+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" file found: `"},"finish_reason":null}]}

409569

410− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Directory"},"finish_reason":null}]}

570+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main.go`\n"},"finish_reason":null}]}

411571

412− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" listing"},"finish_reason":null}]}

572+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"- **LS"},"finish_reason":null}]}

413573

414− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"**: Contains"},"finish_reason":null}]}

574+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"**:"},"finish_reason":null}]}

415575

416− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `go"},"finish_reason":null}]}

576+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Directory contains `"},"finish_reason":null}]}

417577

418− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}]}

578+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"go.mod`"},"finish_reason":null}]}

419579

420− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` and `main"},"finish_reason":null}]}

580+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" and `main"},"finish_reason":null}]}

421581

422− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go`"},"finish_reason":null}]}

582+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go`"},"finish_reason":null}]}

423583

424− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

584+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

425585

426− data: {"id":"chatcmpl-48130e9f88db9c4a09fc17b0","object":"chat.completion.chunk","created":1778597435,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7834,"completion_tokens":33,"total_tokens":7867,"completion_tokens_details":{"reasoning_tokens":32}}}

586+ data: {"id":"chatcmpl-25fded183d7a9936dd525fb3","object":"chat.completion.chunk","created":1778601816,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":863,"completion_tokens":33,"total_tokens":7936,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":32}}}

427587

428588 data: [DONE]

429589

@@ -432,4 +592,4 @@ interactions:

432592 - text/event-stream

433593 status: 200 OK

434594 code: 200

435− duration: 3.926874583s

595+ duration: 13.518372625s

internal/agent/testdata/TestCoderAgent/glm-5.1/read_a_file.yaml+125−237

@@ -25,231 +25,111 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" says"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Read"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" go"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mod"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" should"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Read"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summary"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" go"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mod"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\n"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"think"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\u003e\n\n"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\u003c/"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"think"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\u003e\""},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Lik"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ely"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" we"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Reading"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" produce"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" module"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summar"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"izing"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" this"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"29"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" It's"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" brief"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" request"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" just"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Read"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Reading"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" go"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" the"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" mod"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Go"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" module"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" file"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

131131

132− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" could"},"finish_reason":null}]}

133−

134− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

135−

136− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

137−

138− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Reading"},"finish_reason":null}]}

139−

140− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

141−

142− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" go"},"finish_reason":null}]}

143−

144− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".mod"},"finish_reason":null}]}

145−

146− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

147−

148− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

149−

150− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

151−

152− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

153−

154− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

155−

156− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

157−

158− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

159−

160− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

161−

162− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

163−

164− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

165−

166− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

167−

168− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

169−

170− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

171−

172− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

173−

174− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

175−

176− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

177−

178− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

179−

180− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

181−

182− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

183−

184− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

185−

186− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" One"},"finish_reason":null}]}

187−

188− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

189−

190− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

191−

192− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Provide"},"finish_reason":null}]}

193−

194− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

195−

196− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

197−

198− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Let's"},"finish_reason":null}]}

199−

200− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

201−

202− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

203−

204− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

205−

206− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Reading"},"finish_reason":null}]}

207−

208− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

209−

210− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" go"},"finish_reason":null}]}

211−

212− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".mod"},"finish_reason":null}]}

213−

214− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

215−

216− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

217−

218− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" is"},"finish_reason":null}]}

219−

220− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" about"},"finish_reason":null}]}

221−

222− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

223−

224− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"24"},"finish_reason":null}]}

225−

226− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

227−

228− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

229−

230− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

231−

232− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

233−

234− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Return"},"finish_reason":null}]}

235−

236− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

237−

238− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

239−

240− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Reading"},"finish_reason":null}]}

241−

242− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" the"},"finish_reason":null}]}

243−

244− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" go"},"finish_reason":null}]}

245−

246− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}]}

247−

248− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" file"},"finish_reason":null}]}

249−

250− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

251−

252− data: {"id":"chatcmpl-149388c529739f52423462e3","object":"chat.completion.chunk","created":1778597270,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":15,"completion_tokens":120,"total_tokens":327,"prompt_tokens_details":{"cached_tokens":192}}}

132+ data: {"id":"chatcmpl-34f12044a0a3d5db41d65ef0","object":"chat.completion.chunk","created":1778601665,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":15,"completion_tokens":60,"total_tokens":267,"prompt_tokens_details":{"cached_tokens":192}}}

253133

254134 data: [DONE]

255135

@@ -258,15 +138,15 @@ interactions:

258138 - text/event-stream

259139 status: 200 OK

260140 code: 200

261− duration: 572.800625ms

141+ duration: 882.88625ms

262142- id: 1

263143 request:

264144 proto: HTTP/1.1

265145 proto_major: 1

266146 proto_minor: 1

267− content_length: 33607

147+ content_length: 33694

268148 host: ""

269− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Read the go mod\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

149+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Read the go mod\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

270150 headers:

271151 Accept:

272152 - application/json

@@ -283,43 +163,45 @@ interactions:

283163 content_length: -1

284164 uncompressed: true

285165 body: |+

286− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

287167

288− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

289169

290− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

291171

292− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me"},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me"},"finish_reason":null}]}

293173

294− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to read a"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to read a"},"finish_reason":null}]}

295175

296− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" go.mod"},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" go.mod"},"finish_reason":null}]}

297177

298− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" file. Let"},"finish_reason":null}]}

178+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" file. Let"},"finish_reason":null}]}

299179

300− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me find"},"finish_reason":null}]}

180+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me find"},"finish_reason":null}]}

301181

302− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" it first"},"finish_reason":null}]}

182+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

303183

304− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

184+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" read it"},"finish_reason":null}]}

305185

306− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_6ad0b3d8cf1f49289b0f5fad","type":"function","index":0,"function":{"name":"glob","arguments":""}}]},"finish_reason":null}]}

186+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

307187

308− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

188+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_d7e2b939d2a44b4a8dc06b32","type":"function","index":0,"function":{"name":"glob","arguments":""}}]},"finish_reason":null}]}

309189

310− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

190+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

311191

312− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

313193

314− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"go"}}]},"finish_reason":null}]}

194+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

315195

316− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".mod\""}}]},"finish_reason":null}]}

196+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"**/"}}]},"finish_reason":null}]}

317197

318− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

198+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go.mod\""}}]},"finish_reason":null}]}

319199

320− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

200+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

321201

322− data: {"id":"chatcmpl-793393c6b5de48c25ff680a6","object":"chat.completion.chunk","created":1778597276,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":696,"completion_tokens":29,"total_tokens":7765,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":17}}}

202+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

203+

204+ data: {"id":"chatcmpl-9629fffcdf5950b0d962b784","object":"chat.completion.chunk","created":1778601667,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":721,"completion_tokens":32,"total_tokens":7793,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":18}}}

323205

324206 data: [DONE]

325207

@@ -328,15 +210,15 @@ interactions:

328210 - text/event-stream

329211 status: 200 OK

330212 code: 200

331− duration: 5.728420166s

213+ duration: 3.481540708s

332214- id: 2

333215 request:

334216 proto: HTTP/1.1

335217 proto_major: 1

336218 proto_minor: 1

337− content_length: 33985

219+ content_length: 34078

338220 host: ""

339− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_6ad0b3d8cf1f49289b0f5fad\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"go.mod\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to read a go.mod file. Let me find it first.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file/go.mod\",\"tool_call_id\":\"call_6ad0b3d8cf1f49289b0f5fad\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

221+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_d7e2b939d2a44b4a8dc06b32\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/go.mod\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to read a go.mod file. Let me find and read it.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file/go.mod\",\"tool_call_id\":\"call_d7e2b939d2a44b4a8dc06b32\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

340222 headers:

341223 Accept:

342224 - application/json

@@ -353,43 +235,43 @@ interactions:

353235 content_length: -1

354236 uncompressed: true

355237 body: |+

356− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

357239

358− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_98e2e7b24a214ad685f4a3e5","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_c11e4a0a1d9b46e4b049a005","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

359241

360− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

361243

362− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\"file_path\": "}}]},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

363245

364− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

365247

366− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

367249

368− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

250+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

369251

370− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

252+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

371253

372− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

254+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

373255

374− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

256+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

375257

376− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-"}}]},"finish_reason":null}]}

258+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

377259

378− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5.1/read"}}]},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-"}}]},"finish_reason":null}]}

379261

380− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5.1"}}]},"finish_reason":null}]}

381263

382− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/read_a"}}]},"finish_reason":null}]}

383265

384− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/go.mod"}}]},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}]}

385267

386− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

268+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/go.mod\""}}]},"finish_reason":null}]}

387269

388− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

270+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

389271

390− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

272+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

391273

392− data: {"id":"chatcmpl-bea5d9f86de0fcb4874f50d8","object":"chat.completion.chunk","created":1778597279,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7770,"completion_tokens":29,"total_tokens":7799,"completion_tokens_details":{"reasoning_tokens":28}}}

274+ data: {"id":"chatcmpl-b0780eb38d68c8307b683bae","object":"chat.completion.chunk","created":1778601670,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":53,"completion_tokens":29,"total_tokens":7826,"prompt_tokens_details":{"cached_tokens":7744},"completion_tokens_details":{"reasoning_tokens":28}}}

393275

394276 data: [DONE]

395277

@@ -398,15 +280,15 @@ interactions:

398280 - text/event-stream

399281 status: 200 OK

400282 code: 200

401− duration: 2.767550125s

283+ duration: 1.901558459s

402284- id: 3

403285 request:

404286 proto: HTTP/1.1

405287 proto_major: 1

406288 proto_minor: 1

407− content_length: 34378

289+ content_length: 34471

408290 host: ""

409− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_6ad0b3d8cf1f49289b0f5fad\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"go.mod\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to read a go.mod file. Let me find it first.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file/go.mod\",\"tool_call_id\":\"call_6ad0b3d8cf1f49289b0f5fad\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_98e2e7b24a214ad685f4a3e5\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file/go.mod\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<file>\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n</file>\\n\",\"tool_call_id\":\"call_98e2e7b24a214ad685f4a3e5\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

291+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_d7e2b939d2a44b4a8dc06b32\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/go.mod\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to read a go.mod file. Let me find and read it.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file/go.mod\",\"tool_call_id\":\"call_d7e2b939d2a44b4a8dc06b32\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_c11e4a0a1d9b46e4b049a005\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/read_a_file/go.mod\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<file>\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n</file>\\n\",\"tool_call_id\":\"call_c11e4a0a1d9b46e4b049a005\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

410292 headers:

411293 Accept:

412294 - application/json

@@ -423,37 +305,43 @@ interactions:

423305 content_length: -1

424306 uncompressed: true

425307 body: |+

426− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

308+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

309+

310+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"The"},"finish_reason":null}]}

311+

312+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

313+

314+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}]}

427315

428− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"A"},"finish_reason":null}]}

316+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}]}

429317

430− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" minimal"},"finish_reason":null}]}

318+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` defines"},"finish_reason":null}]}

431319

432− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Go module file"},"finish_reason":null}]}

320+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":\n\n-"},"finish_reason":null}]}

433321

434− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":\n\n- **Module"},"finish_reason":null}]}

322+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" **Module**:"},"finish_reason":null}]}

435323

436− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"**: `"},"finish_reason":null}]}

324+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `example"},"finish_reason":null}]}

437325

438− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"example.com/test"},"finish_reason":null}]}

326+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".com/test"},"finish_reason":null}]}

439327

440− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"project"},"finish_reason":null}]}

328+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"project"},"finish_reason":null}]}

441329

442− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`\n- **"},"finish_reason":null}]}

330+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`\n- **"},"finish_reason":null}]}

443331

444− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Go version"},"finish_reason":null}]}

332+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Go version"},"finish_reason":null}]}

445333

446− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"**: "},"finish_reason":null}]}

334+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"**: "},"finish_reason":null}]}

447335

448− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"1.23"},"finish_reason":null}]}

336+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"1.23"},"finish_reason":null}]}

449337

450− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"\n- No"},"finish_reason":null}]}

338+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}]}

451339

452− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" dependencies"},"finish_reason":null}]}

340+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"- No dependencies"},"finish_reason":null}]}

453341

454− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

342+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

455343

456− data: {"id":"chatcmpl-76b9cbfc7a37c1e9614e4222","object":"chat.completion.chunk","created":1778597282,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7833,"completion_tokens":31,"total_tokens":7864,"completion_tokens_details":{"reasoning_tokens":30}}}

344+ data: {"id":"chatcmpl-d10ef5026cbac5e65b68913f","object":"chat.completion.chunk","created":1778601674,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7860,"completion_tokens":32,"total_tokens":7892,"completion_tokens_details":{"reasoning_tokens":31}}}

457345

458346 data: [DONE]

459347

@@ -462,4 +350,4 @@ interactions:

462350 - text/event-stream

463351 status: 200 OK

464352 code: 200

465− duration: 2.132418458s

353+ duration: 3.047584625s

internal/agent/testdata/TestCoderAgent/glm-5.1/simple_test.yaml+104−64

@@ -25,103 +25,137 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"User"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" asks"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" asks"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\n"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" we"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"think"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\u003e\n\n"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\u003c/"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"think"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summar"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\u003e\""},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"izing"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" is"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" just"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" should"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" should"},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summary"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

122+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-cb802de561d11e076f0b6821","object":"chat.completion.chunk","created":1778597267,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":140,"completion_tokens":56,"total_tokens":260,"prompt_tokens_details":{"cached_tokens":64}}}

124+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

125+

126+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

127+

128+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

129+

130+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" maybe"},"finish_reason":null}]}

131+

132+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

133+

134+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Greeting"},"finish_reason":null}]}

135+

136+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

137+

138+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That"},"finish_reason":null}]}

139+

140+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" is"},"finish_reason":null}]}

141+

142+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

143+

144+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

145+

146+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

147+

148+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" just"},"finish_reason":null}]}

149+

150+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

151+

152+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

153+

154+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Greeting"},"finish_reason":null}]}

155+

156+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

157+

158+ data: {"id":"chatcmpl-9454579d686f7cec9572bb47","object":"chat.completion.chunk","created":1778601659,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":12,"completion_tokens":73,"total_tokens":277,"prompt_tokens_details":{"cached_tokens":192}}}

125159

126160 data: [DONE]

127161

@@ -130,15 +164,15 @@ interactions:

130164 - text/event-stream

131165 status: 200 OK

132166 code: 200

133− duration: 652.235333ms

167+ duration: 793.427542ms

134168- id: 1

135169 request:

136170 proto: HTTP/1.1

137171 proto_major: 1

138172 proto_minor: 1

139− content_length: 33597

173+ content_length: 33684

140174 host: ""

141− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Hello\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

175+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"Hello\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

142176 headers:

143177 Accept:

144178 - application/json

@@ -155,27 +189,33 @@ interactions:

155189 content_length: -1

156190 uncompressed: true

157191 body: |+

158− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

193+

194+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

195+

196+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

197+

198+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" simply"},"finish_reason":null}]}

159199

160− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

200+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" said \""},"finish_reason":null}]}

161201

162− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

202+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Hello"},"finish_reason":null}]}

163203

164− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" said hello"},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"\". I"},"finish_reason":null}]}

165205

166− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":". I"},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"'ll"},"finish_reason":null}]}

167207

168− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" should respond"},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" respond briefly"},"finish_reason":null}]}

169209

170− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" briefly."},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

171211

172− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Hi"},"finish_reason":null}]}

173213

174− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!"},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!"},"finish_reason":null}]}

175215

176− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

216+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

177217

178− data: {"id":"chatcmpl-f5d456fac6115b04b7a65227","object":"chat.completion.chunk","created":1778597269,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7732,"completion_tokens":14,"total_tokens":7746,"completion_tokens_details":{"reasoning_tokens":10}}}

218+ data: {"id":"chatcmpl-54478dac3fa904184b20ee66","object":"chat.completion.chunk","created":1778601664,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7757,"completion_tokens":16,"total_tokens":7773,"completion_tokens_details":{"reasoning_tokens":12}}}

179219

180220 data: [DONE]

181221

@@ -184,4 +224,4 @@ interactions:

184224 - text/event-stream

185225 status: 200 OK

186226 code: 200

187− duration: 3.107841708s

227+ duration: 5.192642417s

internal/agent/testdata/TestCoderAgent/glm-5.1/sourcegraph_tool.yaml+442−518

@@ -25,515 +25,453 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" system"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" says"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" we"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" must"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" generate"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" based"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" on"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user's"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" first"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" message"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" source"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"graph"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" The"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" search"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" says"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"func"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" repositories"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\n"},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"think"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\u003e\""},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" use"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" The"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" source"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"graph"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" search"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" summary"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"func"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" repositories"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" We"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Language"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" is"},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" English"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Not"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Using"},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" more"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Source"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" than"},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"graph"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Find"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" func"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Re"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"pos"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Search"},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" func"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Using"},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Source"},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" repos"},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"graph"},"finish_reason":null}]}

177177

178− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

178+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

179179

180− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Source"},"finish_reason":null}]}

180+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Find"},"finish_reason":null}]}

181181

182− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"graph"},"finish_reason":null}]}

182+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" func"},"finish_reason":null}]}

183183

184− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

184+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

185185

186− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That"},"finish_reason":null}]}

186+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

187187

188− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" is"},"finish_reason":null}]}

188+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

189189

190− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

190+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Re"},"finish_reason":null}]}

191191

192− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"45"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"pos"},"finish_reason":null}]}

193193

194− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

194+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\""},"finish_reason":null}]}

195195

196− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" maybe"},"finish_reason":null}]}

196+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

197197

198− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

198+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

199199

200− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

200+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" let's"},"finish_reason":null}]}

201201

202− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

202+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

203203

204− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

205205

206− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Using"},"finish_reason":null}]}

207207

208− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Search"},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

209209

210− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

211211

212− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" func"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

213213

214− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" +"},"finish_reason":null}]}

215215

216− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

217217

218− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

219219

220− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" repos"},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

221221

222− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

223223

224− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Source"},"finish_reason":null}]}

224+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

225225

226− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"graph"},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

227227

228− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

228+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Source"},"finish_reason":null}]}

229229

230− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

230+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"graph"},"finish_reason":null}]}

231231

232− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

232+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

233233

234− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

234+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"11"},"finish_reason":null}]}

235235

236− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" S"},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

237237

238− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

239239

240− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

241241

242− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

243243

244− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" e"},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

245245

246− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

247247

248− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

249249

250− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"3"},"finish_reason":null}]}

250+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

251251

252− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" r"},"finish_reason":null}]}

252+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"18"},"finish_reason":null}]}

253253

254− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

254+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

255255

256− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" c"},"finish_reason":null}]}

256+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

257257

258− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

258+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

259259

260− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" h"},"finish_reason":null}]}

260+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

261261

262− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

262+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

263263

264− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

264+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"20"},"finish_reason":null}]}

265265

266− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

266+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

267267

268− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

268+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

269269

270− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

270+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

271271

272− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" f"},"finish_reason":null}]}

272+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

273273

274− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"8"},"finish_reason":null}]}

274+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

275275

276− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" o"},"finish_reason":null}]}

276+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"21"},"finish_reason":null}]}

277277

278− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"9"},"finish_reason":null}]}

278+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

279279

280− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" r"},"finish_reason":null}]}

280+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Find"},"finish_reason":null}]}

281281

282− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"10"},"finish_reason":null}]}

282+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

283283

284− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

284+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

285285

286− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

286+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

287287

288− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"11"},"finish_reason":null}]}

288+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"25"},"finish_reason":null}]}

289289

290− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

290+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

291291

292− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" f"},"finish_reason":null}]}

292+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

293293

294− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"12"},"finish_reason":null}]}

294+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

295295

296− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" u"},"finish_reason":null}]}

296+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

297297

298− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"13"},"finish_reason":null}]}

298+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

299299

300− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" n"},"finish_reason":null}]}

300+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"26"},"finish_reason":null}]}

301301

302− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"14"},"finish_reason":null}]}

302+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

303303

304− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" c"},"finish_reason":null}]}

304+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" func"},"finish_reason":null}]}

305305

306− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"15"},"finish_reason":null}]}

306+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

307307

308− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

308+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

309309

310− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

310+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

311311

312− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"16"},"finish_reason":null}]}

312+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"30"},"finish_reason":null}]}

313313

314− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

314+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

315315

316− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" m"},"finish_reason":null}]}

316+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

317317

318− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

318+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

319319

320− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

320+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

321321

322− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"18"},"finish_reason":null}]}

322+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

323323

324− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" i"},"finish_reason":null}]}

324+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"31"},"finish_reason":null}]}

325325

326− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"19"},"finish_reason":null}]}

326+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

327327

328− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" n"},"finish_reason":null}]}

328+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

329329

330− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"20"},"finish_reason":null}]}

330+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

331331

332− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

332+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

333333

334− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

334+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

335335

336− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"21"},"finish_reason":null}]}

336+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"35"},"finish_reason":null}]}

337337

338− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

338+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

339339

340− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" i"},"finish_reason":null}]}

340+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

341341

342− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"22"},"finish_reason":null}]}

342+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

343343

344− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" n"},"finish_reason":null}]}

344+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

345345

346− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"23"},"finish_reason":null}]}

346+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

347347

348− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

348+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"36"},"finish_reason":null}]}

349349

350− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

350+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

351351

352− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"24"},"finish_reason":null}]}

352+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

353353

354− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

354+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

355355

356− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" G"},"finish_reason":null}]}

356+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

357357

358− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"25"},"finish_reason":null}]}

358+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

359359

360− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" o"},"finish_reason":null}]}

360+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

361361

362− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"26"},"finish_reason":null}]}

362+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

363363

364− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

364+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

365365

366− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

366+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

367367

368− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"27"},"finish_reason":null}]}

368+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

369369

370− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

370+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

371371

372− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" r"},"finish_reason":null}]}

372+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"39"},"finish_reason":null}]}

373373

374− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"28"},"finish_reason":null}]}

374+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

375375

376− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" e"},"finish_reason":null}]}

376+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Go"},"finish_reason":null}]}

377377

378− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"29"},"finish_reason":null}]}

378+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

379379

380− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" p"},"finish_reason":null}]}

380+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

381381

382− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"30"},"finish_reason":null}]}

382+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

383383

384− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" o"},"finish_reason":null}]}

384+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"41"},"finish_reason":null}]}

385385

386− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"31"},"finish_reason":null}]}

386+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

387387

388− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" s"},"finish_reason":null}]}

388+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

389389

390− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"32"},"finish_reason":null}]}

390+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

391391

392− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

392+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

393393

394− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

394+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"=\u003e"},"finish_reason":null}]}

395395

396− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"33"},"finish_reason":null}]}

396+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"42"},"finish_reason":null}]}

397397

398− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

398+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

399399

400− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" w"},"finish_reason":null}]}

400+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Re"},"finish_reason":null}]}

401401

402− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"34"},"finish_reason":null}]}

402+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"pos"},"finish_reason":null}]}

403403

404− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" i"},"finish_reason":null}]}

404+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"("},"finish_reason":null}]}

405405

406− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"35"},"finish_reason":null}]}

406+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

407407

408− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" t"},"finish_reason":null}]}

408+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")=\u003e"},"finish_reason":null}]}

409409

410− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"36"},"finish_reason":null}]}

410+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"47"},"finish_reason":null}]}

411411

412− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" h"},"finish_reason":null}]}

412+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

413413

414− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"37"},"finish_reason":null}]}

414+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Under"},"finish_reason":null}]}

415415

416− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

416+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

417417

418− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"space"},"finish_reason":null}]}

418+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

419419

420− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

420+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

421421

422− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

422+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

423423

424− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" S"},"finish_reason":null}]}

424+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

425425

426− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"39"},"finish_reason":null}]}

426+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

427427

428− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" o"},"finish_reason":null}]}

428+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

429429

430− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"40"},"finish_reason":null}]}

430+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

431431

432− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" u"},"finish_reason":null}]}

432+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

433433

434− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"41"},"finish_reason":null}]}

434+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

435435

436− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" r"},"finish_reason":null}]}

436+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

437437

438− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"42"},"finish_reason":null}]}

438+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" One"},"finish_reason":null}]}

439439

440− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" c"},"finish_reason":null}]}

440+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

441441

442− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"43"},"finish_reason":null}]}

442+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

443443

444− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" e"},"finish_reason":null}]}

444+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

445445

446− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"44"},"finish_reason":null}]}

446+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

447447

448− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" g"},"finish_reason":null}]}

448+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

449449

450− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"45"},"finish_reason":null}]}

450+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Using"},"finish_reason":null}]}

451451

452− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" r"},"finish_reason":null}]}

452+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Source"},"finish_reason":null}]}

453453

454− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"46"},"finish_reason":null}]}

454+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"graph"},"finish_reason":null}]}

455455

456− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

456+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

457457

458− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"47"},"finish_reason":null}]}

458+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Find"},"finish_reason":null}]}

459459

460− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" p"},"finish_reason":null}]}

460+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" func"},"finish_reason":null}]}

461461

462− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"48"},"finish_reason":null}]}

462+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}]}

463463

464− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" h"},"finish_reason":null}]}

464+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

465465

466− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"49"},"finish_reason":null}]}

466+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Go"},"finish_reason":null}]}

467467

468− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

468+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Re"},"finish_reason":null}]}

469469

470− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

470+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"pos"},"finish_reason":null}]}

471471

472− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

472+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

473473

474− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"49"},"finish_reason":null}]}

475−

476− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

477−

478− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

479−

480− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" fine"},"finish_reason":null}]}

481−

482− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" (\u003c"},"finish_reason":null}]}

483−

484− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

485−

486− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

487−

488− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

489−

490− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

491−

492− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

493−

494− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" colon"},"finish_reason":null}]}

495−

496− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

497−

498− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

499−

500− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

501−

502− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

503−

504− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

505−

506− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

507−

508− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Return"},"finish_reason":null}]}

509−

510− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

511−

512− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

513−

514− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Search"},"finish_reason":null}]}

515−

516− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" for"},"finish_reason":null}]}

517−

518− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" func"},"finish_reason":null}]}

519−

520− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}]}

521−

522− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

523−

524− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Go"},"finish_reason":null}]}

525−

526− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" repos"},"finish_reason":null}]}

527−

528− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" with"},"finish_reason":null}]}

529−

530− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Source"},"finish_reason":null}]}

531−

532− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"graph"},"finish_reason":null}]}

533−

534− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

535−

536− data: {"id":"chatcmpl-35bd80f4ca07eeaee979b148","object":"chat.completion.chunk","created":1778597398,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":152,"completion_tokens":262,"total_tokens":478,"prompt_tokens_details":{"cached_tokens":64}}}

474+ data: {"id":"chatcmpl-b7b08c82ee26f8d6d7cb91fc","object":"chat.completion.chunk","created":1778601767,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":24,"completion_tokens":231,"total_tokens":447,"prompt_tokens_details":{"cached_tokens":192}}}

537475

538476 data: [DONE]

539477

@@ -542,15 +480,15 @@ interactions:

542480 - text/event-stream

543481 status: 200 OK

544482 code: 200

545− duration: 620.468833ms

483+ duration: 432.945167ms

546484- id: 1

547485 request:

548486 proto: HTTP/1.1

549487 proto_major: 1

550488 proto_minor: 1

551− content_length: 33657

489+ content_length: 33744

552490 host: ""

553− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

491+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

554492 headers:

555493 Accept:

556494 - application/json

@@ -567,53 +505,55 @@ interactions:

567505 content_length: -1

568506 uncompressed: true

569507 body: |+

570− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

508+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

509+

510+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

571511

572− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

512+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

573513

574− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

514+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me to"},"finish_reason":null}]}

575515

576− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me to"},"finish_reason":null}]}

516+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" search"},"finish_reason":null}]}

577517

578− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" use the"},"finish_reason":null}]}

518+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" for '"},"finish_reason":null}]}

579519

580− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" Sourcegraph"},"finish_reason":null}]}

520+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"func"},"finish_reason":null}]}

581521

582− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" tool to"},"finish_reason":null}]}

522+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" main' in"},"finish_reason":null}]}

583523

584− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" search for '"},"finish_reason":null}]}

524+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" Go repositories"},"finish_reason":null}]}

585525

586− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"func"},"finish_reason":null}]}

526+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" using the"},"finish_reason":null}]}

587527

588− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" main' in"},"finish_reason":null}]}

528+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" Sourcegraph"},"finish_reason":null}]}

589529

590− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" Go repositories"},"finish_reason":null}]}

530+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" tool."},"finish_reason":null}]}

591531

592− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

532+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_2119903f18294120853e87c3","type":"function","index":0,"function":{"name":"sourcegraph","arguments":""}}]},"finish_reason":null}]}

593533

594− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_3377a986e2324af1a0d22066","type":"function","index":0,"function":{"name":"sourcegraph","arguments":""}}]},"finish_reason":null}]}

534+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

595535

596− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

536+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

597537

598− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

538+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"count\": "}}]},"finish_reason":null}]}

599539

600− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"count\": "}}]},"finish_reason":null}]}

540+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"20"}}]},"finish_reason":null}]}

601541

602− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"20"}}]},"finish_reason":null}]}

542+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

603543

604− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", \"query\": "}}]},"finish_reason":null}]}

544+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"query\": "}}]},"finish_reason":null}]}

605545

606− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"func main"}}]},"finish_reason":null}]}

546+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"func main"}}]},"finish_reason":null}]}

607547

608− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" lang:Go"}}]},"finish_reason":null}]}

548+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" lang:"}}]},"finish_reason":null}]}

609549

610− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

550+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go\""}}]},"finish_reason":null}]}

611551

612− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

552+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

613553

614− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

554+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

615555

616− data: {"id":"chatcmpl-adbd389de8021f205d5cc62b","object":"chat.completion.chunk","created":1778597402,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7745,"completion_tokens":43,"total_tokens":7788,"completion_tokens_details":{"reasoning_tokens":21}}}

556+ data: {"id":"chatcmpl-1507a9a06fbf24ffeba5605d","object":"chat.completion.chunk","created":1778601770,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":730,"completion_tokens":42,"total_tokens":7812,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":20}}}

617557

618558 data: [DONE]

619559

@@ -622,7 +562,7 @@ interactions:

622562 - text/event-stream

623563 status: 200 OK

624564 code: 200

625− duration: 3.692503542s

565+ duration: 3.269446209s

626566- id: 2

627567 request:

628568 proto: HTTP/1.1

@@ -630,7 +570,7 @@ interactions:

630570 proto_minor: 1

631571 content_length: 424

632572 host: sourcegraph.com

633− body: '{"query":"query Search($query: String!) { search(query: $query, version: V2, patternType: keyword ) { results { matchCount, limitHit, resultCount, approximateResultCount, missing { name }, timedout { name }, indexUnavailable, results { __typename, ... on FileMatch { repository { name }, file { path, url, content }, lineMatches { preview, lineNumber, offsetAndLengths } } } } } }","variables":{"query":"func main lang:Go"}}'

573+ body: '{"query":"query Search($query: String!) { search(query: $query, version: V2, patternType: keyword ) { results { matchCount, limitHit, resultCount, approximateResultCount, missing { name }, timedout { name }, indexUnavailable, results { __typename, ... on FileMatch { repository { name }, file { path, url, content }, lineMatches { preview, lineNumber, offsetAndLengths } } } } } }","variables":{"query":"func main lang:go"}}'

634574 headers:

635575 Content-Type:

636576 - application/json

@@ -644,21 +584,21 @@ interactions:

644584 proto_minor: 0

645585 content_length: -1

646586 uncompressed: true

647− body: '{"data":{"search":{"results":{"matchCount":30,"limitHit":true,"resultCount":30,"approximateResultCount":"30+","missing":[],"timedout":[],"indexUnavailable":false,"results":[{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/01_before-func-expression/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc greeting() {\n\tfmt.Println(\"Hello world!\")\n}\n\nfunc main() {\n\tgreeting()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":8,"offsetAndLengths":[[0,9]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func greeting() {","lineNumber":4,"offsetAndLengths":[[0,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/02_func-expression/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc main() {\n\n\tgreeting := func() {\n\t\tfmt.Println(\"Hello world!\")\n\t}\n\n\tgreeting()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":4,"offsetAndLengths":[[0,9]]},{"preview":"\tgreeting := func() {","lineNumber":6,"offsetAndLengths":[[13,4]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/03_func-expression_shows-type/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/03_func-expression_shows-type/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc main() {\n\n\tgreeting := func() {\n\t\tfmt.Println(\"Hello world!\")\n\t}\n\n\tgreeting()\n\tfmt.Printf(\"%T\\n\", greeting)\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":4,"offsetAndLengths":[[0,9]]},{"preview":"\tgreeting := func() {","lineNumber":6,"offsetAndLengths":[[13,4]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/04_another-way_func-expression/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/04_another-way_func-expression/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc makeGreeter() func() string {\n\treturn func() string {\n\t\treturn \"Hello world!\"\n\t}\n}\n\nfunc main() {\n\tgreet := makeGreeter()\n\tfmt.Println(greet())\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":10,"offsetAndLengths":[[0,9]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func makeGreeter() func() string {","lineNumber":4,"offsetAndLengths":[[0,4],[19,4]]},{"preview":"\treturn func() string {","lineNumber":5,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc makeGreeter() func() string {\n\treturn func() string {\n\t\treturn \"Hello world!\"\n\t}\n}\n\nfunc main() {\n\tgreet := makeGreeter()\n\tfmt.Println(greet())\n\tfmt.Printf(\"%T\\n\", greet)\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":10,"offsetAndLengths":[[0,9]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func makeGreeter() func() string {","lineNumber":4,"offsetAndLengths":[[0,4],[19,4]]},{"preview":"\treturn func() string {","lineNumber":5,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/12_callbacks/01_print-nums/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/12_callbacks/01_print-nums/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc visit(numbers []int, callback func(int)) {\n\tfor _, n := range numbers {\n\t\tcallback(n)\n\t}\n}\n\nfunc main() {\n\tvisit([]int{1, 2, 3, 4}, func(n int) {\n\t\tfmt.Println(n)\n\t})\n}\n\n// callback: passing a func as an argument\n"},"lineMatches":[{"preview":"func main() {","lineNumber":10,"offsetAndLengths":[[0,9]]},{"preview":"\tvisit([]int{1, 2, 3, 4}, func(n int) {","lineNumber":11,"offsetAndLengths":[[26,4]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func visit(numbers []int, callback func(int)) {","lineNumber":4,"offsetAndLengths":[[0,4],[35,4]]},{"preview":"// callback: passing a func as an argument","lineNumber":16,"offsetAndLengths":[[23,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/14_defer/01_no-defer/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/14_defer/01_no-defer/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc hello() {\n\tfmt.Print(\"hello \")\n}\n\nfunc world() {\n\tfmt.Println(\"world\")\n}\n\nfunc main() {\n\tworld()\n\thello()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":12,"offsetAndLengths":[[0,9]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func hello() {","lineNumber":4,"offsetAndLengths":[[0,4]]},{"preview":"func world() {","lineNumber":8,"offsetAndLengths":[[0,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/14_defer/02_with-defer/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/14_defer/02_with-defer/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc hello() {\n\tfmt.Print(\"hello \")\n}\n\nfunc world() {\n\tfmt.Println(\"world\")\n}\n\nfunc main() {\n\tdefer world()\n\thello()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":12,"offsetAndLengths":[[0,9]]}]}]}}}}'

587+ body: '{"data":{"search":{"results":{"matchCount":30,"limitHit":true,"resultCount":30,"approximateResultCount":"30+","missing":[],"timedout":[],"indexUnavailable":false,"results":[{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/01_before-func-expression/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc greeting() {\n\tfmt.Println(\"Hello world!\")\n}\n\nfunc main() {\n\tgreeting()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":8,"offsetAndLengths":[[0,9]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func greeting() {","lineNumber":4,"offsetAndLengths":[[0,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/02_func-expression/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc main() {\n\n\tgreeting := func() {\n\t\tfmt.Println(\"Hello world!\")\n\t}\n\n\tgreeting()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":4,"offsetAndLengths":[[0,9]]},{"preview":"\tgreeting := func() {","lineNumber":6,"offsetAndLengths":[[13,4]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/03_func-expression_shows-type/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/03_func-expression_shows-type/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc main() {\n\n\tgreeting := func() {\n\t\tfmt.Println(\"Hello world!\")\n\t}\n\n\tgreeting()\n\tfmt.Printf(\"%T\\n\", greeting)\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":4,"offsetAndLengths":[[0,9]]},{"preview":"\tgreeting := func() {","lineNumber":6,"offsetAndLengths":[[13,4]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/04_another-way_func-expression/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/04_another-way_func-expression/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc makeGreeter() func() string {\n\treturn func() string {\n\t\treturn \"Hello world!\"\n\t}\n}\n\nfunc main() {\n\tgreet := makeGreeter()\n\tfmt.Println(greet())\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":10,"offsetAndLengths":[[0,9]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func makeGreeter() func() string {","lineNumber":4,"offsetAndLengths":[[0,4],[19,4]]},{"preview":"\treturn func() string {","lineNumber":5,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/GoesToEleven/GolangTraining"},"file":{"path":"14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go","url":"/r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc makeGreeter() func() string {\n\treturn func() string {\n\t\treturn \"Hello world!\"\n\t}\n}\n\nfunc main() {\n\tgreet := makeGreeter()\n\tfmt.Println(greet())\n\tfmt.Printf(\"%T\\n\", greet)\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":10,"offsetAndLengths":[[0,9]]},{"preview":"package main","lineNumber":0,"offsetAndLengths":[[8,4]]},{"preview":"func makeGreeter() func() string {","lineNumber":4,"offsetAndLengths":[[0,4],[19,4]]},{"preview":"\treturn func() string {","lineNumber":5,"offsetAndLengths":[[8,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/mauricioabreu/golings"},"file":{"path":"exercises/functions/functions1/main.go","url":"/r/github.com/mauricioabreu/golings/-/blob/exercises/functions/functions1/main.go","content":"// functions1\n// Make me compile!\n\n// I AM NOT DONE\npackage main\n\nfunc main() {\n\tcall_me()\n}\n"},"lineMatches":[{"preview":"package main","lineNumber":4,"offsetAndLengths":[[8,4]]},{"preview":"func main() {","lineNumber":6,"offsetAndLengths":[[0,9]]},{"preview":"// functions1","lineNumber":0,"offsetAndLengths":[[3,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/mauricioabreu/golings"},"file":{"path":"exercises/anonymous_functions/anonymous_functions2/main.go","url":"/r/github.com/mauricioabreu/golings/-/blob/exercises/anonymous_functions/anonymous_functions2/main.go","content":"// anonymous functions2\n// Make me compile!\n\n// I AM NOT DONE\npackage main\n\nimport \"fmt\"\n\nfunc main() {\n\tvar sayBye func(name string)\n\n\tsayBye = func() {\n\t\tfmt.Printf(\"Bye %s\", n)\n\t}\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":8,"offsetAndLengths":[[0,9]]},{"preview":"\tvar sayBye func(name string)","lineNumber":9,"offsetAndLengths":[[12,4]]},{"preview":"\tsayBye = func() {","lineNumber":11,"offsetAndLengths":[[10,4]]},{"preview":"package main","lineNumber":4,"offsetAndLengths":[[8,4]]},{"preview":"// anonymous functions2","lineNumber":0,"offsetAndLengths":[[13,4]]}]},{"__typename":"FileMatch","repository":{"name":"github.com/mauricioabreu/golings"},"file":{"path":"exercises/anonymous_functions/anonymous_functions1/main.go","url":"/r/github.com/mauricioabreu/golings/-/blob/exercises/anonymous_functions/anonymous_functions1/main.go","content":"// anonymous functions1\n// Make me compile!\n\n// I AM NOT DONE\npackage main\n\nimport \"fmt\"\n\nfunc main() {\n\n\tfunc(name string) {\n\t\tfmt.Printf(\"Hello %s\", name)\n\t}()\n\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":8,"offsetAndLengths":[[0,9]]},{"preview":"\tfunc(name string) {","lineNumber":10,"offsetAndLengths":[[1,4]]},{"preview":"package main","lineNumber":4,"offsetAndLengths":[[8,4]]}]}]}}}}'

648588 headers:

649589 Content-Type:

650590 - application/json

651591 status: 200 OK

652592 code: 200

653− duration: 2.251490458s

593+ duration: 1.383383209s

654594- id: 3

655595 request:

656596 proto: HTTP/1.1

657597 proto_major: 1

658598 proto_minor: 1

659− content_length: 42734

599+ content_length: 42747

660600 host: ""

661− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_3377a986e2324af1a0d22066\",\"function\":{\"arguments\":\"{\\\"count\\\": 20, \\\"query\\\": \\\"func main lang:Go\\\"}\",\"name\":\"sourcegraph\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to use the Sourcegraph tool to search for 'func main' in Go repositories.\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| func main() {\\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func greeting() {\\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n## Result 2: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/02_func-expression/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 3: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/03_func-expression_shows-type/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/03_func-expression_shows-type/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| \\tfmt.Printf(\\\"%T\\\\n\\\", greeting)\\n13| }\\n14| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| \\tfmt.Printf(\\\"%T\\\\n\\\", greeting)\\n13| }\\n14| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 4: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/04_another-way_func-expression/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/04_another-way_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| func main() {\\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| }\\n15| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func makeGreeter() func() string {\\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| }\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\treturn func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| }\\n15| \\n```\\n\\n## Result 5: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| func main() {\\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| \\tfmt.Printf(\\\"%T\\\\n\\\", greet)\\n15| }\\n16| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func makeGreeter() func() string {\\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| \\tfmt.Printf(\\\"%T\\\\n\\\", greet)\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\treturn func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| \\tfmt.Printf(\\\"%T\\\\n\\\", greet)\\n15| }\\n```\\n\\n## Result 6: github.com/GoesToEleven/GolangTraining/14_functions/12_callbacks/01_print-nums/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/12_callbacks/01_print-nums/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func visit(numbers []int, callback func(int)) {\\n6| \\tfor _, n := range numbers {\\n7| \\t\\tcallback(n)\\n8| \\t}\\n9| }\\n10| func main() {\\n11| func main() {\\n12| \\tvisit([]int{1, 2, 3, 4}, func(n int) {\\n13| \\t\\tfmt.Println(n)\\n14| \\t})\\n15| }\\n16| \\n17| // callback: passing a func as an argument\\n18| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func visit(numbers []int, callback func(int)) {\\n6| \\tfor _, n := range numbers {\\n7| \\t\\tcallback(n)\\n8| \\t}\\n9| }\\n10| \\n11| \\tvisit([]int{1, 2, 3, 4}, func(n int) {\\n12| \\tvisit([]int{1, 2, 3, 4}, func(n int) {\\n13| \\t\\tfmt.Println(n)\\n14| \\t})\\n15| }\\n16| \\n17| // callback: passing a func as an argument\\n18| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func visit(numbers []int, callback func(int)) {\\n6| \\tfor _, n := range numbers {\\n7| \\t\\tcallback(n)\\n8| \\t}\\n9| }\\n10| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func visit(numbers []int, callback func(int)) {\\n5| func visit(numbers []int, callback func(int)) {\\n6| \\tfor _, n := range numbers {\\n7| \\t\\tcallback(n)\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tvisit([]int{1, 2, 3, 4}, func(n int) {\\n13| \\t\\tfmt.Println(n)\\n14| \\t})\\n```\\n\\n```\\n6| \\tfor _, n := range numbers {\\n7| \\t\\tcallback(n)\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tvisit([]int{1, 2, 3, 4}, func(n int) {\\n13| \\t\\tfmt.Println(n)\\n14| \\t})\\n15| }\\n16| // callback: passing a func as an argument\\n17| // callback: passing a func as an argument\\n18| \\n```\\n\\n## Result 7: github.com/GoesToEleven/GolangTraining/14_functions/14_defer/01_no-defer/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/14_defer/01_no-defer/main.go\\n\\n```\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func hello() {\\n6| \\tfmt.Print(\\\"hello \\\")\\n7| }\\n8| \\n9| func world() {\\n10| \\tfmt.Println(\\\"world\\\")\\n11| }\\n12| func main() {\\n13| func main() {\\n14| \\tworld()\\n15| \\thello()\\n16| }\\n17| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func hello() {\\n6| \\tfmt.Print(\\\"hello \\\")\\n7| }\\n8| \\n9| func world() {\\n10| \\tfmt.Println(\\\"world\\\")\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func hello() {\\n5| func hello() {\\n6| \\tfmt.Print(\\\"hello \\\")\\n7| }\\n8| \\n9| func world() {\\n10| \\tfmt.Println(\\\"world\\\")\\n11| }\\n12| \\n13| func main() {\\n14| \\tworld()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func hello() {\\n6| \\tfmt.Print(\\\"hello \\\")\\n7| }\\n8| func world() {\\n9| func world() {\\n10| \\tfmt.Println(\\\"world\\\")\\n11| }\\n12| \\n13| func main() {\\n14| \\tworld()\\n15| \\thello()\\n16| }\\n17| \\n```\\n\\n## Result 8: github.com/GoesToEleven/GolangTraining/14_functions/14_defer/02_with-defer/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/14_defer/02_with-defer/main.go\\n\\n```\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func hello() {\\n6| \\tfmt.Print(\\\"hello \\\")\\n7| }\\n8| \\n9| func world() {\\n10| \\tfmt.Println(\\\"world\\\")\\n11| }\\n12| func main() {\\n13| func main() {\\n14| \\tdefer world()\\n15| \\thello()\\n16| }\\n17| \\n```\\n\\n\",\"tool_call_id\":\"call_3377a986e2324af1a0d22066\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

601+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_2119903f18294120853e87c3\",\"function\":{\"arguments\":\"{\\\"count\\\": 20, \\\"query\\\": \\\"func main lang:go\\\"}\",\"name\":\"sourcegraph\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to search for 'func main' in Go repositories using the Sourcegraph tool.\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| func main() {\\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func greeting() {\\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n## Result 2: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/02_func-expression/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 3: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/03_func-expression_shows-type/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/03_func-expression_shows-type/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| \\tfmt.Printf(\\\"%T\\\\n\\\", greeting)\\n13| }\\n14| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| \\tfmt.Printf(\\\"%T\\\\n\\\", greeting)\\n13| }\\n14| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 4: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/04_another-way_func-expression/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/04_another-way_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| func main() {\\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| }\\n15| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func makeGreeter() func() string {\\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| }\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\treturn func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| }\\n15| \\n```\\n\\n## Result 5: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go\\n\\nURL: /r/github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/05_another-way_func-expression_shows-type/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| func main() {\\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| \\tfmt.Printf(\\\"%T\\\\n\\\", greet)\\n15| }\\n16| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func makeGreeter() func() string {\\n5| func makeGreeter() func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| \\tfmt.Printf(\\\"%T\\\\n\\\", greet)\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\treturn func() string {\\n6| \\treturn func() string {\\n7| \\t\\treturn \\\"Hello world!\\\"\\n8| \\t}\\n9| }\\n10| \\n11| func main() {\\n12| \\tgreet := makeGreeter()\\n13| \\tfmt.Println(greet())\\n14| \\tfmt.Printf(\\\"%T\\\\n\\\", greet)\\n15| }\\n```\\n\\n## Result 6: github.com/mauricioabreu/golings/exercises/functions/functions1/main.go\\n\\nURL: /r/github.com/mauricioabreu/golings/-/blob/exercises/functions/functions1/main.go\\n\\n```\\n1| // functions1\\n2| // Make me compile!\\n3| \\n4| package main\\n5| package main\\n6| \\n7| func main() {\\n8| \\tcall_me()\\n9| }\\n10| \\n```\\n\\n```\\n1| // functions1\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| func main() {\\n7| func main() {\\n8| \\tcall_me()\\n9| }\\n10| \\n```\\n\\n```\\n0| // functions1\\n1| // functions1\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| \\n7| func main() {\\n8| \\tcall_me()\\n9| }\\n10| \\n```\\n\\n## Result 7: github.com/mauricioabreu/golings/exercises/anonymous_functions/anonymous_functions2/main.go\\n\\nURL: /r/github.com/mauricioabreu/golings/-/blob/exercises/anonymous_functions/anonymous_functions2/main.go\\n\\n```\\n1| // anonymous functions2\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| func main() {\\n9| func main() {\\n10| \\tvar sayBye func(name string)\\n11| \\n12| \\tsayBye = func() {\\n13| \\t\\tfmt.Printf(\\\"Bye %s\\\", n)\\n14| \\t}\\n15| }\\n16| \\n```\\n\\n```\\n1| // anonymous functions2\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| \\n9| \\tvar sayBye func(name string)\\n10| \\tvar sayBye func(name string)\\n11| \\n12| \\tsayBye = func() {\\n13| \\t\\tfmt.Printf(\\\"Bye %s\\\", n)\\n14| \\t}\\n15| }\\n16| \\n```\\n\\n```\\n1| // anonymous functions2\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| \\n9| func main() {\\n10| \\tvar sayBye func(name string)\\n11| \\tsayBye = func() {\\n12| \\tsayBye = func() {\\n13| \\t\\tfmt.Printf(\\\"Bye %s\\\", n)\\n14| \\t}\\n15| }\\n16| \\n```\\n\\n```\\n1| // anonymous functions2\\n2| // Make me compile!\\n3| \\n4| package main\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| \\n9| func main() {\\n10| \\tvar sayBye func(name string)\\n11| \\n12| \\tsayBye = func() {\\n13| \\t\\tfmt.Printf(\\\"Bye %s\\\", n)\\n14| \\t}\\n```\\n\\n```\\n0| // anonymous functions2\\n1| // anonymous functions2\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| \\n9| func main() {\\n10| \\tvar sayBye func(name string)\\n```\\n\\n## Result 8: github.com/mauricioabreu/golings/exercises/anonymous_functions/anonymous_functions1/main.go\\n\\nURL: /r/github.com/mauricioabreu/golings/-/blob/exercises/anonymous_functions/anonymous_functions1/main.go\\n\\n```\\n1| // anonymous functions1\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| func main() {\\n9| func main() {\\n10| \\n11| \\tfunc(name string) {\\n12| \\t\\tfmt.Printf(\\\"Hello %s\\\", name)\\n13| \\t}()\\n14| \\n15| }\\n16| \\n```\\n\\n```\\n1| // anonymous functions1\\n2| // Make me compile!\\n3| \\n4| // I AM NOT DONE\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| \\n9| func main() {\\n10| \\tfunc(name string) {\\n11| \\tfunc(name string) {\\n12| \\t\\tfmt.Printf(\\\"Hello %s\\\", name)\\n13| \\t}()\\n14| \\n15| }\\n16| \\n```\\n\\n```\\n1| // anonymous functions1\\n2| // Make me compile!\\n3| \\n4| package main\\n5| package main\\n6| \\n7| import \\\"fmt\\\"\\n8| \\n9| func main() {\\n10| \\n11| \\tfunc(name string) {\\n12| \\t\\tfmt.Printf(\\\"Hello %s\\\", name)\\n13| \\t}()\\n14| \\n```\\n\\n\",\"tool_call_id\":\"call_2119903f18294120853e87c3\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

662602 headers:

663603 Accept:

664604 - application/json

@@ -675,385 +615,369 @@ interactions:

675615 content_length: -1

676616 uncompressed: true

677617 body: |+

678− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

679−

680− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Found"},"finish_reason":null}]}

681−

682− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" **"},"finish_reason":null}]}

683−

684− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"30 matches"},"finish_reason":null}]}

685−

686− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"** across Go"},"finish_reason":null}]}

687−

688− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" repositories. Here"},"finish_reason":null}]}

689−

690− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"'s"},"finish_reason":null}]}

691−

692− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" a summary of the"},"finish_reason":null}]}

693−

694− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" top results"},"finish_reason":null}]}

618+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

695619

696− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":\n\n| #"},"finish_reason":null}]}

620+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Found"},"finish_reason":null}]}

697621

698− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | Repository"},"finish_reason":null}]}

622+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 30"},"finish_reason":null}]}

699623

700− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | File"},"finish_reason":null}]}

624+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"+ matches"},"finish_reason":null}]}

701625

702− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | Description"},"finish_reason":null}]}

626+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" across Go"},"finish_reason":null}]}

703627

704− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" |\n|---"},"finish_reason":null}]}

628+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" repositories"},"finish_reason":null}]}

705629

706− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"|-----------"},"finish_reason":null}]}

630+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":". Here"},"finish_reason":null}]}

707631

708− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"|------"},"finish_reason":null}]}

632+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"'s a summary"},"finish_reason":null}]}

709633

710− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"|-------------"},"finish_reason":null}]}

634+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" of the"},"finish_reason":null}]}

711635

712− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"|\n| "},"finish_reason":null}]}

636+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" top results"},"finish_reason":null}]}

713637

714− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"1 | `G"},"finish_reason":null}]}

638+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":\n\n|"},"finish_reason":null}]}

715639

716− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesTo"},"finish_reason":null}]}

640+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" #"},"finish_reason":null}]}

717641

718− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Ele"},"finish_reason":null}]}

642+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | Repository"},"finish_reason":null}]}

719643

720− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/G"},"finish_reason":null}]}

644+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | File"},"finish_reason":null}]}

721645

722− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olangTraining"},"finish_reason":null}]}

646+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | Context"},"finish_reason":null}]}

723647

724− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | `"},"finish_reason":null}]}

648+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" |\n|"},"finish_reason":null}]}

725649

726− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"14_functions"},"finish_reason":null}]}

650+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"---|-----------"},"finish_reason":null}]}

727651

728− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/10"},"finish_reason":null}]}

652+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"|------"},"finish_reason":null}]}

729653

730− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func"},"finish_reason":null}]}

654+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"|---------"},"finish_reason":null}]}

731655

732− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression"},"finish_reason":null}]}

656+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"|\n|"},"finish_reason":null}]}

733657

734− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}]}

658+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 1 | `"},"finish_reason":null}]}

735659

736− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"01_before"},"finish_reason":null}]}

660+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Goes"},"finish_reason":null}]}

737661

738− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-f"},"finish_reason":null}]}

662+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"To"},"finish_reason":null}]}

739663

740− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"unc-expression"},"finish_reason":null}]}

664+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Ele"},"finish_reason":null}]}

741665

742− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}]}

666+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven"},"finish_reason":null}]}

743667

744− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}]}

668+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/Golang"},"finish_reason":null}]}

745669

746− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | Basic"},"finish_reason":null}]}

670+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Training"},"finish_reason":null}]}

747671

748− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" function"},"finish_reason":null}]}

672+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | `..."},"finish_reason":null}]}

749673

750− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" call"},"finish_reason":null}]}

674+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/01"},"finish_reason":null}]}

751675

752− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" from"},"finish_reason":null}]}

676+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_before"},"finish_reason":null}]}

753677

754− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

678+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-f"},"finish_reason":null}]}

755679

756− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main"},"finish_reason":null}]}

680+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"unc-expression"},"finish_reason":null}]}

757681

758− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |\n"},"finish_reason":null}]}

682+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go"},"finish_reason":null}]}

759683

760− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"| "},"finish_reason":null}]}

684+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | Simple"},"finish_reason":null}]}

761685

762− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"2 | `G"},"finish_reason":null}]}

686+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

763687

764− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

688+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"func"},"finish_reason":null}]}

765689

766− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/GolangTraining"},"finish_reason":null}]}

690+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" main()`"},"finish_reason":null}]}

767691

768− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |"},"finish_reason":null}]}

692+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" calling `g"},"finish_reason":null}]}

769693

770− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

694+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"reeting()`"},"finish_reason":null}]}

771695

772− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"14_functions/10"},"finish_reason":null}]}

696+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" |\n|"},"finish_reason":null}]}

773697

774− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func-expression/02"},"finish_reason":null}]}

698+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 2"},"finish_reason":null}]}

775699

776− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func"},"finish_reason":null}]}

700+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | `G"},"finish_reason":null}]}

777701

778− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression/main.go"},"finish_reason":null}]}

702+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

779703

780− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | Function"},"finish_reason":null}]}

704+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/Golang"},"finish_reason":null}]}

781705

782− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" expression assigned"},"finish_reason":null}]}

706+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Training` |"},"finish_reason":null}]}

783707

784− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

708+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `.../02"},"finish_reason":null}]}

785709

786− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `main"},"finish_reason":null}]}

710+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func"},"finish_reason":null}]}

787711

788− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |\n| "},"finish_reason":null}]}

712+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression"},"finish_reason":null}]}

789713

790− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"3"},"finish_reason":null}]}

714+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go"},"finish_reason":null}]}

791715

792− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | `G"},"finish_reason":null}]}

716+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | `func"},"finish_reason":null}]}

793717

794− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

718+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" main()`"},"finish_reason":null}]}

795719

796− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/GolangTraining"},"finish_reason":null}]}

720+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" with a"},"finish_reason":null}]}

797721

798− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |"},"finish_reason":null}]}

722+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" function expression"},"finish_reason":null}]}

799723

800− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `14_functions/"},"finish_reason":null}]}

724+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" assigned to a"},"finish_reason":null}]}

801725

802− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"10_func-expression/"},"finish_reason":null}]}

726+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" variable |\n"},"finish_reason":null}]}

803727

804− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"03_func"},"finish_reason":null}]}

728+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"| "},"finish_reason":null}]}

805729

806− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression_sh"},"finish_reason":null}]}

730+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"3 | `G"},"finish_reason":null}]}

807731

808− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ows-type"},"finish_reason":null}]}

732+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

809733

810− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go` |"},"finish_reason":null}]}

734+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/Golang"},"finish_reason":null}]}

811735

812− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Function"},"finish_reason":null}]}

736+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Training` |"},"finish_reason":null}]}

813737

814− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" expression with"},"finish_reason":null}]}

738+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `.../"},"finish_reason":null}]}

815739

816− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" type printing"},"finish_reason":null}]}

740+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"03_func"},"finish_reason":null}]}

817741

818− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" |\n| "},"finish_reason":null}]}

742+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression_sh"},"finish_reason":null}]}

819743

820− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"4 | `"},"finish_reason":null}]}

744+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ows-type"},"finish_reason":null}]}

821745

822− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"GoesToEle"},"finish_reason":null}]}

746+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go"},"finish_reason":null}]}

823747

824− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/Golang"},"finish_reason":null}]}

748+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | Same"},"finish_reason":null}]}

825749

826− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Training` |"},"finish_reason":null}]}

750+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":","},"finish_reason":null}]}

827751

828− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `14_functions/"},"finish_reason":null}]}

752+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" plus"},"finish_reason":null}]}

829753

830− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"10_func-expression/"},"finish_reason":null}]}

754+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" printing the"},"finish_reason":null}]}

831755

832− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"04_an"},"finish_reason":null}]}

756+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" type |\n"},"finish_reason":null}]}

833757

834− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"other"},"finish_reason":null}]}

758+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"| "},"finish_reason":null}]}

835759

836− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-way"},"finish_reason":null}]}

760+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"4 | `G"},"finish_reason":null}]}

837761

838− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func"},"finish_reason":null}]}

762+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

839763

840− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression/main.go"},"finish_reason":null}]}

764+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/GolangTraining"},"finish_reason":null}]}

841765

842− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | Function"},"finish_reason":null}]}

766+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |"},"finish_reason":null}]}

843767

844− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" returning"},"finish_reason":null}]}

768+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `.../04"},"finish_reason":null}]}

845769

846− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" a"},"finish_reason":null}]}

770+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_another-way"},"finish_reason":null}]}

847771

848− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" closure |\n"},"finish_reason":null}]}

772+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func"},"finish_reason":null}]}

849773

850− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"| "},"finish_reason":null}]}

774+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression/main.go`"},"finish_reason":null}]}

851775

852− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"5 | `G"},"finish_reason":null}]}

776+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | `func"},"finish_reason":null}]}

853777

854− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

778+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" main()` using"},"finish_reason":null}]}

855779

856− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/G"},"finish_reason":null}]}

780+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" a closure"},"finish_reason":null}]}

857781

858− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olangTraining` |"},"finish_reason":null}]}

782+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" returned from `make"},"finish_reason":null}]}

859783

860− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `14"},"finish_reason":null}]}

784+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Gre"},"finish_reason":null}]}

861785

862− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_functions/10_func"},"finish_reason":null}]}

786+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"eter"},"finish_reason":null}]}

863787

864− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-expression/05"},"finish_reason":null}]}

788+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"()` |\n|"},"finish_reason":null}]}

865789

866− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_an"},"finish_reason":null}]}

790+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 5 |"},"finish_reason":null}]}

867791

868− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"other-way"},"finish_reason":null}]}

792+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `G"},"finish_reason":null}]}

869793

870− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func-expression_sh"},"finish_reason":null}]}

794+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

871795

872− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ows-type"},"finish_reason":null}]}

796+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/G"},"finish_reason":null}]}

873797

874− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go` |"},"finish_reason":null}]}

798+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olangTraining"},"finish_reason":null}]}

875799

876− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Closure with type"},"finish_reason":null}]}

800+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |"},"finish_reason":null}]}

877801

878− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" inspection |\n"},"finish_reason":null}]}

802+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `..."},"finish_reason":null}]}

879803

880− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"| "},"finish_reason":null}]}

804+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/05_an"},"finish_reason":null}]}

881805

882− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"6 | `G"},"finish_reason":null}]}

806+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"other-way"},"finish_reason":null}]}

883807

884− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

808+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_func-expression"},"finish_reason":null}]}

885809

886− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/Golang"},"finish_reason":null}]}

810+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_shows-type"},"finish_reason":null}]}

887811

888− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Training` |"},"finish_reason":null}]}

812+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go"},"finish_reason":null}]}

889813

890− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `14"},"finish_reason":null}]}

814+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |"},"finish_reason":null}]}

891815

892− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_functions/12"},"finish_reason":null}]}

816+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Same with"},"finish_reason":null}]}

893817

894− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_callbacks"},"finish_reason":null}]}

818+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" type"},"finish_reason":null}]}

895819

896− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/01"},"finish_reason":null}]}

820+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" inspection"},"finish_reason":null}]}

897821

898− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_print"},"finish_reason":null}]}

822+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" |\n|"},"finish_reason":null}]}

899823

900− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}]}

824+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 6 |"},"finish_reason":null}]}

901825

902− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"nums/main.go`"},"finish_reason":null}]}

826+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `maur"},"finish_reason":null}]}

903827

904− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | Callback pattern"},"finish_reason":null}]}

828+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"icioab"},"finish_reason":null}]}

905829

906− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" in `main"},"finish_reason":null}]}

830+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"reu"},"finish_reason":null}]}

907831

908− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` |\n|"},"finish_reason":null}]}

832+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/g"},"finish_reason":null}]}

909833

910− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 7 | `"},"finish_reason":null}]}

834+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olings"},"finish_reason":null}]}

911835

912− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"GoesToEle"},"finish_reason":null}]}

836+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | `"},"finish_reason":null}]}

913837

914− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/G"},"finish_reason":null}]}

838+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ex"},"finish_reason":null}]}

915839

916− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olangTraining` |"},"finish_reason":null}]}

840+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ercises/functions"},"finish_reason":null}]}

917841

918− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `14"},"finish_reason":null}]}

842+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/functions"},"finish_reason":null}]}

919843

920− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_functions/14"},"finish_reason":null}]}

844+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"1"},"finish_reason":null}]}

921845

922− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_def"},"finish_reason":null}]}

846+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go` |"},"finish_reason":null}]}

923847

924− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"er/"},"finish_reason":null}]}

848+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Exercise"},"finish_reason":null}]}

925849

926− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"01_no"},"finish_reason":null}]}

850+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}]}

927851

928− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-defer/main"},"finish_reason":null}]}

852+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" calling"},"finish_reason":null}]}

929853

930− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go` |"},"finish_reason":null}]}

854+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `call"},"finish_reason":null}]}

931855

932− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Simple"},"finish_reason":null}]}

856+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_me()`"},"finish_reason":null}]}

933857

934− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

858+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" from"},"finish_reason":null}]}

935859

936− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main` without"},"finish_reason":null}]}

860+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" main |\n|"},"finish_reason":null}]}

937861

938− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

862+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" 7 |"},"finish_reason":null}]}

939863

940− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"defer`"},"finish_reason":null}]}

864+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `mauricio"},"finish_reason":null}]}

941865

942− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" |\n| "},"finish_reason":null}]}

866+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"abreu"},"finish_reason":null}]}

943867

944− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"8"},"finish_reason":null}]}

868+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/golings`"},"finish_reason":null}]}

945869

946− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | `G"},"finish_reason":null}]}

870+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | `ex"},"finish_reason":null}]}

947871

948− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

872+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ercises/anonymous_functions"},"finish_reason":null}]}

949873

950− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/G"},"finish_reason":null}]}

874+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/anonymous"},"finish_reason":null}]}

951875

952− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olangTraining` |"},"finish_reason":null}]}

876+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_functions2"},"finish_reason":null}]}

953877

954− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `14_functions/"},"finish_reason":null}]}

878+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go` |"},"finish_reason":null}]}

955879

956− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"14_defer/"},"finish_reason":null}]}

880+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Exercise: anonymous"},"finish_reason":null}]}

957881

958− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"02_with-def"},"finish_reason":null}]}

882+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" function assignment |\n"},"finish_reason":null}]}

959883

960− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"er/main.go"},"finish_reason":null}]}

884+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"| "},"finish_reason":null}]}

961885

962− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` | `main"},"finish_reason":null}]}

886+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"8 |"},"finish_reason":null}]}

963887

964− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` using"},"finish_reason":null}]}

888+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `mauricio"},"finish_reason":null}]}

965889

966− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `defer`"},"finish_reason":null}]}

890+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"abreu"},"finish_reason":null}]}

967891

968− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" |\n\nMost"},"finish_reason":null}]}

892+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/golings`"},"finish_reason":null}]}

969893

970− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" results are from"},"finish_reason":null}]}

894+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" | `ex"},"finish_reason":null}]}

971895

972− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" the **"},"finish_reason":null}]}

896+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ercises/anonymous"},"finish_reason":null}]}

973897

974− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"GoesToEle"},"finish_reason":null}]}

898+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_functions/anonymous"},"finish_reason":null}]}

975899

976− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/G"},"finish_reason":null}]}

900+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"_functions1"},"finish_reason":null}]}

977901

978− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olangTraining** tutorial"},"finish_reason":null}]}

902+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/main.go` |"},"finish_reason":null}]}

979903

980− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" repo"},"finish_reason":null}]}

904+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" Exercise: immediately"},"finish_reason":null}]}

981905

982− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":". To"},"finish_reason":null}]}

906+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"-inv"},"finish_reason":null}]}

983907

984− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" narrow"},"finish_reason":null}]}

908+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oked anonymous"},"finish_reason":null}]}

985909

986− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" results to"},"finish_reason":null}]}

910+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" function |\n\nThe"},"finish_reason":null}]}

987911

988− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" more"},"finish_reason":null}]}

912+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" results"},"finish_reason":null}]}

989913

990− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" diverse"},"finish_reason":null}]}

914+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" are primarily"},"finish_reason":null}]}

991915

992− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}]}

916+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" from Go"},"finish_reason":null}]}

993917

994− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"production"},"finish_reason":null}]}

918+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" training/tutorial"},"finish_reason":null}]}

995919

996− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" code"},"finish_reason":null}]}

920+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" repos"},"finish_reason":null}]}

997921

998− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":", you"},"finish_reason":null}]}

922+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" (`Golang"},"finish_reason":null}]}

999923

1000− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" could refine the"},"finish_reason":null}]}

924+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Training`)"},"finish_reason":null}]}

1001925

1002− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" query,"},"finish_reason":null}]}

926+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" and exercise"},"finish_reason":null}]}

1003927

1004− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" e.g.:\n\n"},"finish_reason":null}]}

928+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" collections"},"finish_reason":null}]}

1005929

1006− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"- `func"},"finish_reason":null}]}

930+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" (`g"},"finish_reason":null}]}

1007931

1008− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" main lang"},"finish_reason":null}]}

932+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olings`). To"},"finish_reason":null}]}

1009933

1010− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}]}

934+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" narrow"},"finish_reason":null}]}

1011935

1012− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Go -"},"finish_reason":null}]}

936+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" results to"},"finish_reason":null}]}

1013937

1014− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"repo:"},"finish_reason":null}]}

938+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" production"},"finish_reason":null}]}

1015939

1016− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"G"},"finish_reason":null}]}

940+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" code"},"finish_reason":null}]}

1017941

1018− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oesToEle"},"finish_reason":null}]}

942+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":", you"},"finish_reason":null}]}

1019943

1020− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven/GolangTraining"},"finish_reason":null}]}

944+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" could add repo"},"finish_reason":null}]}

1021945

1022− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` ("},"finish_reason":null}]}

946+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" filters"},"finish_reason":null}]}

1023947

1024− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"exclude"},"finish_reason":null}]}

948+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" ("},"finish_reason":null}]}

1025949

1026− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" training"},"finish_reason":null}]}

950+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"e.g"},"finish_reason":null}]}

1027951

1028− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" repo)\n-"},"finish_reason":null}]}

952+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"., `-"},"finish_reason":null}]}

1029953

1030− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `func main lang"},"finish_reason":null}]}

954+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"repo"},"finish_reason":null}]}

1031955

1032− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":Go count"},"finish_reason":null}]}

956+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":G"},"finish_reason":null}]}

1033957

1034− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}]}

958+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"oes"},"finish_reason":null}]}

1035959

1036− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"50` ("},"finish_reason":null}]}

960+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ToEle"},"finish_reason":null}]}

1037961

1038− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"more"},"finish_reason":null}]}

962+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"ven -"},"finish_reason":null}]}

1039963

1040− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" results)\n"},"finish_reason":null}]}

964+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"repo:ma"},"finish_reason":null}]}

1041965

1042− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"- `func"},"finish_reason":null}]}

966+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"uricio"},"finish_reason":null}]}

1043967

1044− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" main lang:Go"},"finish_reason":null}]}

968+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"abreu`)"},"finish_reason":null}]}

1045969

1046− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" repo:g"},"finish_reason":null}]}

970+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" or search"},"finish_reason":null}]}

1047971

1048− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"olang` (standard"},"finish_reason":null}]}

972+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" for more"},"finish_reason":null}]}

1049973

1050− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" library only"},"finish_reason":null}]}

974+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" specific patterns"},"finish_reason":null}]}

1051975

1052− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":")"},"finish_reason":null}]}

976+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"."},"finish_reason":null}]}

1053977

1054− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

978+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

1055979

1056− data: {"id":"chatcmpl-b1b4d4b3a19eaa32ad806633","object":"chat.completion.chunk","created":1778597408,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":10586,"completion_tokens":444,"total_tokens":11030,"completion_tokens_details":{"reasoning_tokens":443}}}

980+ data: {"id":"chatcmpl-886ba4df675175dc008d9a5e","object":"chat.completion.chunk","created":1778601779,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":10569,"completion_tokens":397,"total_tokens":10966,"completion_tokens_details":{"reasoning_tokens":396}}}

1057981

1058982 data: [DONE]

1059983

@@ -1062,4 +986,4 @@ interactions:

1062986 - text/event-stream

1063987 status: 200 OK

1064988 code: 200

1065− duration: 3.460100417s

989+ duration: 6.201306583s

internal/agent/testdata/TestCoderAgent/glm-5.1/update_a_file.yaml+157−453

@@ -25,391 +25,145 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"We"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" need"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" wants"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" output"},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" given"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"update"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" by"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Update"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" changing"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" print"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" print"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" say"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"hello"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" crush"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" crush"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'\""},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" must"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Something"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"English"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" like"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")."},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Must"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Update"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" print"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" hello"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" crush"},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" within"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" or"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" col"},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Possible"},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Return"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Modify"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" just"},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" that"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" print"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Update"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" print"},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Count"},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" hello"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" from"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" crush"},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

165165

166− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Modify"},"finish_reason":null}]}

167−

168− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

169−

170− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

171−

172− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

173−

174− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" print"},"finish_reason":null}]}

175−

176− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

177−

178− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

179−

180− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

181−

182− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

183−

184− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

185−

186− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

187−

188− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

189−

190− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Modify"},"finish_reason":null}]}

191−

192− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

193−

194− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"6"},"finish_reason":null}]}

195−

196− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

197−

198− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" +"},"finish_reason":null}]}

199−

200− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

201−

202− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

203−

204− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

205−

206− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

207−

208− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ="},"finish_reason":null}]}

209−

210− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

211−

212− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

213−

214− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

215−

216− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

217−

218− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

219−

220− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"7"},"finish_reason":null}]}

221−

222− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")"},"finish_reason":null}]}

223−

224− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" =\u003e"},"finish_reason":null}]}

225−

226− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" total"},"finish_reason":null}]}

227−

228− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

229−

230− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"14"},"finish_reason":null}]}

231−

232− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

233−

234− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

235−

236− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

237−

238− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

239−

240− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

241−

242− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"15"},"finish_reason":null}]}

243−

244− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

245−

246− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

247−

248− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

249−

250− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"2"},"finish_reason":null}]}

251−

252− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

253−

254− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"17"},"finish_reason":null}]}

255−

256− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

257−

258− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

259−

260− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

261−

262− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"18"},"finish_reason":null}]}

263−

264− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

265−

266− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" print"},"finish_reason":null}]}

267−

268− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

269−

270− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

271−

272− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

273−

274− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"23"},"finish_reason":null}]}

275−

276− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

277−

278− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

279−

280− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

281−

282− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"24"},"finish_reason":null}]}

283−

284− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

285−

286− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

287−

288− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

289−

290− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

291−

292− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

293−

294− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"29"},"finish_reason":null}]}

295−

296− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

297−

298− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

299−

300− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

301−

302− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"30"},"finish_reason":null}]}

303−

304− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

305−

306− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

307−

308− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

309−

310− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"4"},"finish_reason":null}]}

311−

312− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

313−

314− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"34"},"finish_reason":null}]}

315−

316− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

317−

318− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" space"},"finish_reason":null}]}

319−

320− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"="},"finish_reason":null}]}

321−

322− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"35"},"finish_reason":null}]}

323−

324− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

325−

326− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

327−

328− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ("},"finish_reason":null}]}

329−

330− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"5"},"finish_reason":null}]}

331−

332− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":")="},"finish_reason":null}]}

333−

334− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"40"},"finish_reason":null}]}

335−

336− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

337−

338− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

339−

340− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" length"},"finish_reason":null}]}

341−

342− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

343−

344− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"40"},"finish_reason":null}]}

345−

346− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

347−

348− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" okay"},"finish_reason":null}]}

349−

350− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

351−

352− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Alternatively"},"finish_reason":null}]}

353−

354− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

355−

356− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

357−

358− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Update"},"finish_reason":null}]}

359−

360− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" main"},"finish_reason":null}]}

361−

362− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

363−

364− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

365−

366− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" say"},"finish_reason":null}]}

367−

368− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Hello"},"finish_reason":null}]}

369−

370− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" from"},"finish_reason":null}]}

371−

372− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Crush"},"finish_reason":null}]}

373−

374− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

375−

376− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

377−

378− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" similar"},"finish_reason":null}]}

379−

380− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" length"},"finish_reason":null}]}

381−

382− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

383−

384− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"I'll"},"finish_reason":null}]}

385−

386− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" output"},"finish_reason":null}]}

387−

388− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

389−

390− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

391−

392− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

393−

394− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Update"},"finish_reason":null}]}

395−

396− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}]}

397−

398− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}]}

399−

400− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}]}

401−

402− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" print"},"finish_reason":null}]}

403−

404− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Hello"},"finish_reason":null}]}

405−

406− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" from"},"finish_reason":null}]}

407−

408− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" Crush"},"finish_reason":null}]}

409−

410− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

411−

412− data: {"id":"chatcmpl-27f6fc9e843256fb939896e2","object":"chat.completion.chunk","created":1778597283,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":25,"completion_tokens":200,"total_tokens":417,"prompt_tokens_details":{"cached_tokens":192}}}

166+ data: {"id":"chatcmpl-b190cf1ebc9ab9733b9f1f4c","object":"chat.completion.chunk","created":1778601675,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":37,"completion_tokens":77,"total_tokens":290,"prompt_tokens_details":{"cached_tokens":176}}}

413167

414168 data: [DONE]

415169

@@ -418,15 +172,15 @@ interactions:

418172 - text/event-stream

419173 status: 200 OK

420174 code: 200

421− duration: 628.654542ms

175+ duration: 516.61525ms

422176- id: 1

423177 request:

424178 proto: HTTP/1.1

425179 proto_major: 1

426180 proto_minor: 1

427− content_length: 33663

181+ content_length: 33750

428182 host: ""

429− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

183+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

430184 headers:

431185 Accept:

432186 - application/json

@@ -443,113 +197,57 @@ interactions:

443197 content_length: -1

444198 uncompressed: true

445199 body: |+

446− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

200+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

447201

448− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Let"},"finish_reason":null}]}

202+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Let"},"finish_reason":null}]}

449203

450− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me"},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" me"},"finish_reason":null}]}

451205

452− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" find"},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" read"},"finish_reason":null}]}

453207

454− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" the main"},"finish_reason":null}]}

455209

456− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" read the main"},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":".go"},"finish_reason":null}]}

457211

458− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":".go file"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" file first"},"finish_reason":null}]}

459213

460− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" first."},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

461215

462− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_336a65a4f1974f07938cae66","type":"function","index":0,"function":{"name":"glob","arguments":""}}]},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_c7972dbf84cf4d40996ea0f5","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

463217

464− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

218+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

465219

466− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

220+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

467221

468− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"pattern\": "}}]},"finish_reason":null}]}

222+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": \"/tmp"}}]},"finish_reason":null}]}

469223

470− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"**/"}}]},"finish_reason":null}]}

224+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

471225

472− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"main.go\""}}]},"finish_reason":null}]}

226+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

473227

474− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

228+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

475229

476− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

230+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

477231

478− data: {"id":"chatcmpl-c8c07f692210879bd28e2e82","object":"chat.completion.chunk","created":1778597286,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7746,"completion_tokens":25,"total_tokens":7771,"completion_tokens_details":{"reasoning_tokens":11}}}

232+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"CoderAgent"}}]},"finish_reason":null}]}

479233

480− data: [DONE]

234+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

481235

482− headers:

483− Content-Type:

484− - text/event-stream

485− status: 200 OK

486− code: 200

487− duration: 3.702887459s

488−- id: 2

489− request:

490− proto: HTTP/1.1

491− proto_major: 1

492− proto_minor: 1

493− content_length: 34030

494− host: ""

495− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_336a65a4f1974f07938cae66\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/main.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me find and read the main.go file first.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\",\"tool_call_id\":\"call_336a65a4f1974f07938cae66\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

496− headers:

497− Accept:

498− - application/json

499− Content-Type:

500− - application/json

501− User-Agent:

502− - Charm-Crush/devel (https://charm.land/crush)

503− url: https://hyper.charm.land/v1/chat/completions

504− method: POST

505− response:

506− proto: HTTP/2.0

507− proto_major: 2

508− proto_minor: 0

509− content_length: -1

510− uncompressed: true

511− body: |+

512− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

513−

514− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_009a902e80d94744baddc43c","type":"function","index":0,"function":{"name":"view","arguments":""}}]},"finish_reason":null}]}

515−

516− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

236+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

517237

518− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

238+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1"}}]},"finish_reason":null}]}

519239

520− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

240+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update_a"}}]},"finish_reason":null}]}

521241

522− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

242+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}]}

523243

524− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

244+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go\""}}]},"finish_reason":null}]}

525245

526− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

246+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

527247

528− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

248+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

529249

530− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

531−

532− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/g"}}]},"finish_reason":null}]}

533−

534− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lm-"}}]},"finish_reason":null}]}

535−

536− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5."}}]},"finish_reason":null}]}

537−

538− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1/update"}}]},"finish_reason":null}]}

539−

540− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}]}

541−

542− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}]}

543−

544− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go"}}]},"finish_reason":null}]}

545−

546− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

547−

548− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

549−

550− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

551−

552− data: {"id":"chatcmpl-d376241c8483befb37b6c5bc","object":"chat.completion.chunk","created":1778597297,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7782,"completion_tokens":29,"total_tokens":7811,"completion_tokens_details":{"reasoning_tokens":28}}}

250+ data: {"id":"chatcmpl-0757ac94af53c3fb25b4b78d","object":"chat.completion.chunk","created":1778601682,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7771,"completion_tokens":38,"total_tokens":7809,"completion_tokens_details":{"reasoning_tokens":9}}}

553251

554252 data: [DONE]

555253

@@ -558,15 +256,15 @@ interactions:

558256 - text/event-stream

559257 status: 200 OK

560258 code: 200

561− duration: 10.221187583s

562−- id: 3

259+ duration: 7.904001708s

260+- id: 2

563261 request:

564262 proto: HTTP/1.1

565263 proto_major: 1

566264 proto_minor: 1

567− content_length: 34497

265+ content_length: 34252

568266 host: ""

569− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_336a65a4f1974f07938cae66\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/main.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me find and read the main.go file first.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\",\"tool_call_id\":\"call_336a65a4f1974f07938cae66\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_009a902e80d94744baddc43c\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_009a902e80d94744baddc43c\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

267+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_c7972dbf84cf4d40996ea0f5\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me read the main.go file first.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_c7972dbf84cf4d40996ea0f5\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

570268 headers:

571269 Accept:

572270 - application/json

@@ -583,63 +281,59 @@ interactions:

583281 content_length: -1

584282 uncompressed: true

585283 body: |+

586− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

587−

588− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_37b8cae848074063b7be4bd2","type":"function","index":0,"function":{"name":"edit","arguments":""}}]},"finish_reason":null}]}

284+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

589285

590− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

286+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_2621f1453a094bdabc79a754","type":"function","index":0,"function":{"name":"edit","arguments":""}}]},"finish_reason":null}]}

591287

592− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

288+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

593289

594− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": \"/tmp"}}]},"finish_reason":null}]}

290+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

595291

596− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/crush"}}]},"finish_reason":null}]}

292+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

597293

598− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

294+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp"}}]},"finish_reason":null}]}

599295

600− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

296+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/c"}}]},"finish_reason":null}]}

601297

602− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

298+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

603299

604− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

300+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

605301

606− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

302+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"CoderAgent"}}]},"finish_reason":null}]}

607303

608− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

304+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm-"}}]},"finish_reason":null}]}

609305

610− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1"}}]},"finish_reason":null}]}

306+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5.1/update"}}]},"finish_reason":null}]}

611307

612− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update_a"}}]},"finish_reason":null}]}

308+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a_file"}}]},"finish_reason":null}]}

613309

614− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}]}

310+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go\""}}]},"finish_reason":null}]}

615311

616− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main.go"}}]},"finish_reason":null}]}

312+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", "}}]},"finish_reason":null}]}

617313

618− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

314+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"new_string\": "}}]},"finish_reason":null}]}

619315

620− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"new_string\": "}}]},"finish_reason":null}]}

316+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"\\tfmt.Println(\\\""}}]},"finish_reason":null}]}

621317

622− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"fmt"}}]},"finish_reason":null}]}

318+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello from crush"}}]},"finish_reason":null}]}

623319

624− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println(\\\"Hello"}}]},"finish_reason":null}]}

320+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")"}}]},"finish_reason":null}]}

625321

626− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" from crush"}}]},"finish_reason":null}]}

322+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\", "}}]},"finish_reason":null}]}

627323

628− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")\", "}}]},"finish_reason":null}]}

324+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"old_string\": "}}]},"finish_reason":null}]}

629325

630− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"old_string\": "}}]},"finish_reason":null}]}

326+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"\\tfmt.Println(\\\"Hello"}}]},"finish_reason":null}]}

631327

632− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"fmt.Println(\\\"Hello"}}]},"finish_reason":null}]}

328+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", World"}}]},"finish_reason":null}]}

633329

634− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":", World"}}]},"finish_reason":null}]}

330+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")\""}}]},"finish_reason":null}]}

635331

636− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\\")\""}}]},"finish_reason":null}]}

332+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

637333

638− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

334+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

639335

640− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

641−

642− data: {"id":"chatcmpl-24a95c8d63629662a3c0d8c8","object":"chat.completion.chunk","created":1778597303,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":123,"completion_tokens":55,"total_tokens":7922,"prompt_tokens_details":{"cached_tokens":7744},"completion_tokens_details":{"reasoning_tokens":54}}}

336+ data: {"id":"chatcmpl-93ffe58970d74aec43ed0e13","object":"chat.completion.chunk","created":1778601686,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":7856,"completion_tokens":55,"total_tokens":7911,"completion_tokens_details":{"reasoning_tokens":54}}}

643337

644338 data: [DONE]

645339

@@ -648,15 +342,15 @@ interactions:

648342 - text/event-stream

649343 status: 200 OK

650344 code: 200

651− duration: 5.29761375s

652−- id: 4

345+ duration: 2.8873885s

346+- id: 3

653347 request:

654348 proto: HTTP/1.1

655349 proto_major: 1

656350 proto_minor: 1

657− content_length: 35037

351+ content_length: 34798

658352 host: ""

659− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_336a65a4f1974f07938cae66\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/main.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me find and read the main.go file first.\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\",\"tool_call_id\":\"call_336a65a4f1974f07938cae66\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_009a902e80d94744baddc43c\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_009a902e80d94744baddc43c\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_37b8cae848074063b7be4bd2\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\\", \\\"new_string\\\": \\\"fmt.Println(\\\\\\\"Hello from crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"fmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\n</result>\\n\",\"tool_call_id\":\"call_37b8cae848074063b7be4bd2\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

353+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_c7972dbf84cf4d40996ea0f5\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Let me read the main.go file first.\"},{\"content\":\"<file>\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n</file>\\n\",\"tool_call_id\":\"call_c7972dbf84cf4d40996ea0f5\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_2621f1453a094bdabc79a754\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\\", \\\"new_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello from crush!\\\\\\\")\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"\"},{\"content\":\"<result>\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/glm-5.1/update_a_file/main.go\\n</result>\\n\",\"tool_call_id\":\"call_2621f1453a094bdabc79a754\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

660354 headers:

661355 Accept:

662356 - application/json

@@ -673,29 +367,39 @@ interactions:

673367 content_length: -1

674368 uncompressed: true

675369 body: |+

676− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

370+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

371+

372+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}]}

373+

374+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":". Changed the print"},"finish_reason":null}]}

375+

376+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" message"},"finish_reason":null}]}

377+

378+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" in"},"finish_reason":null}]}

379+

380+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

677381

678− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}]}

382+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"main.go:6"},"finish_reason":null}]}

679383

680− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" —"},"finish_reason":null}]}

384+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`"},"finish_reason":null}]}

681385

682− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" updated `main"},"finish_reason":null}]}

386+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" from"},"finish_reason":null}]}

683387

684− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}]}

388+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `\""},"finish_reason":null}]}

685389

686− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":":6"},"finish_reason":null}]}

390+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Hello, World"},"finish_reason":null}]}

687391

688− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"` to print"},"finish_reason":null}]}

392+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!\"` to `\""},"finish_reason":null}]}

689393

690− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `\"Hello"},"finish_reason":null}]}

394+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Hello from"},"finish_reason":null}]}

691395

692− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" from crush"},"finish_reason":null}]}

396+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" crush!\""},"finish_reason":null}]}

693397

694− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"!\"`."},"finish_reason":null}]}

398+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`."},"finish_reason":null}]}

695399

696− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

400+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

697401

698− data: {"id":"chatcmpl-31f9d9d9ccc6ba0c5f4124ec","object":"chat.completion.chunk","created":1778597308,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":149,"completion_tokens":19,"total_tokens":7976,"prompt_tokens_details":{"cached_tokens":7808},"completion_tokens_details":{"reasoning_tokens":18}}}

402+ data: {"id":"chatcmpl-1dc6071e8f1300c6c1cacfae","object":"chat.completion.chunk","created":1778601690,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":138,"completion_tokens":29,"total_tokens":7975,"prompt_tokens_details":{"cached_tokens":7808},"completion_tokens_details":{"reasoning_tokens":28}}}

699403

700404 data: [DONE]

701405

@@ -704,4 +408,4 @@ interactions:

704408 - text/event-stream

705409 status: 200 OK

706410 code: 200

707− duration: 4.751869542s

411+ duration: 3.548258583s

internal/agent/testdata/TestCoderAgent/glm-5.1/write_tool.yaml+228−156

@@ -25,199 +25,289 @@ interactions:

2525 content_length: -1

2626 uncompressed: true

2727 body: |+

28− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

28+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

2929

30− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

30+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

3131

32− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

32+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

3333

34− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" asks"},"finish_reason":null}]}

34+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" asks"},"finish_reason":null}]}

3535

36− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

36+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

3737

38− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

38+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

3939

40− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

40+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Generate"},"finish_reason":null}]}

4141

42− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

42+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

4343

44− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

44+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" concise"},"finish_reason":null}]}

4545

46− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

46+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

4747

48− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

48+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" for"},"finish_reason":null}]}

4949

50− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

50+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" the"},"finish_reason":null}]}

5151

52− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

52+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" following"},"finish_reason":null}]}

5353

54− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

54+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

5555

56− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":\n\n"},"finish_reason":null}]}

56+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

5757

58− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"use"},"finish_reason":null}]}

58+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" use"},"finish_reason":null}]}

5959

60− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" write"},"finish_reason":null}]}

60+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" write"},"finish_reason":null}]}

6161

62− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

62+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" to"},"finish_reason":null}]}

6363

64− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

64+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" create"},"finish_reason":null}]}

6565

66− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

66+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

6767

68− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" new"},"finish_reason":null}]}

68+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" new"},"finish_reason":null}]}

6969

70− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

70+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

7171

72− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" called"},"finish_reason":null}]}

72+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" called"},"finish_reason":null}]}

7373

74− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" config"},"finish_reason":null}]}

74+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" config"},"finish_reason":null}]}

7575

76− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".json"},"finish_reason":null}]}

76+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".json"},"finish_reason":null}]}

7777

78− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

78+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

7979

80− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

80+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

8181

82− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '{\""},"finish_reason":null}]}

82+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" '{\""},"finish_reason":null}]}

8383

84− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"name"},"finish_reason":null}]}

84+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"name"},"finish_reason":null}]}

8585

86− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\":"},"finish_reason":null}]}

86+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\":"},"finish_reason":null}]}

8787

88− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

88+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

8989

90− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"test"},"finish_reason":null}]}

90+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"test"},"finish_reason":null}]}

9191

92− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\","},"finish_reason":null}]}

92+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\","},"finish_reason":null}]}

9393

94− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

94+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

9595

96− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"version"},"finish_reason":null}]}

96+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"version"},"finish_reason":null}]}

9797

98− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\":"},"finish_reason":null}]}

98+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\":"},"finish_reason":null}]}

9999

100− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

100+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

101101

102− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

102+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"1"},"finish_reason":null}]}

103103

104− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

104+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

105105

106− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"0"},"finish_reason":null}]}

106+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"0"},"finish_reason":null}]}

107107

108− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

108+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

109109

110− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"0"},"finish_reason":null}]}

110+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"0"},"finish_reason":null}]}

111111

112− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"}"},"finish_reason":null}]}

112+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"}"},"finish_reason":null}]}

113113

114− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

114+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"'"},"finish_reason":null}]}

115115

116− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

116+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

117117

118− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Need"},"finish_reason":null}]}

118+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" So"},"finish_reason":null}]}

119119

120− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

120+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" output"},"finish_reason":null}]}

121121

122− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" short"},"finish_reason":null}]}

122+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" a"},"finish_reason":null}]}

123123

124− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

124+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

125125

126− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \u003c="},"finish_reason":null}]}

126+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

127127

128− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

128+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

129129

130− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

130+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

131131

132− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

132+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

133133

134− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

134+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

135135

136− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

136+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" one"},"finish_reason":null}]}

137137

138− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

138+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" line"},"finish_reason":null}]}

139139

140− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

140+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":","},"finish_reason":null}]}

141141

142− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" colon"},"finish_reason":null}]}

142+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" no"},"finish_reason":null}]}

143143

144− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

144+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

145145

146− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Summ"},"finish_reason":null}]}

146+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"/"},"finish_reason":null}]}

147147

148− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ar"},"finish_reason":null}]}

148+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"col"},"finish_reason":null}]}

149149

150− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ize"},"finish_reason":null}]}

150+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"ons"},"finish_reason":null}]}

151151

152− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

152+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

153153

154− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

154+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

155155

156− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Create"},"finish_reason":null}]}

156+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

157157

158− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" config"},"finish_reason":null}]}

158+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" something"},"finish_reason":null}]}

159159

160− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".json"},"finish_reason":null}]}

160+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" like"},"finish_reason":null}]}

161161

162− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" file"},"finish_reason":null}]}

162+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

163163

164− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

164+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Create"},"finish_reason":null}]}

165165

166− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" JSON"},"finish_reason":null}]}

166+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" config"},"finish_reason":null}]}

167167

168− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

168+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".json"},"finish_reason":null}]}

169169

170− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

170+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

171171

172− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Title"},"finish_reason":null}]}

172+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

173173

174− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" length"},"finish_reason":null}]}

174+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" settings"},"finish_reason":null}]}

175175

176− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" ~"},"finish_reason":null}]}

176+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

177177

178− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"38"},"finish_reason":null}]}

178+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

179179

180− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

180+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" under"},"finish_reason":null}]}

181181

182− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

182+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

183183

184− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

184+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"50"},"finish_reason":null}]}

185185

186− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

186+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

187187

188− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That"},"finish_reason":null}]}

188+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

189189

190− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" works"},"finish_reason":null}]}

190+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

191191

192− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".\n\n"},"finish_reason":null}]}

192+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

193193

194− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Return"},"finish_reason":null}]}

194+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Should"},"finish_reason":null}]}

195195

196− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" only"},"finish_reason":null}]}

196+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" be"},"finish_reason":null}]}

197197

198− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" title"},"finish_reason":null}]}

198+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" in"},"finish_reason":null}]}

199199

200− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

200+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" English"},"finish_reason":null}]}

201201

202− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Create"},"finish_reason":null}]}

202+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" same"},"finish_reason":null}]}

203203

204− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" config"},"finish_reason":null}]}

204+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" language"},"finish_reason":null}]}

205205

206− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".json"},"finish_reason":null}]}

206+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

207207

208− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" file"},"finish_reason":null}]}

208+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" We'll"},"finish_reason":null}]}

209209

210− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" with"},"finish_reason":null}]}

210+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" output"},"finish_reason":null}]}

211211

212− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" specified"},"finish_reason":null}]}

212+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

213213

214− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" JSON"},"finish_reason":null}]}

214+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Create"},"finish_reason":null}]}

215215

216− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" content"},"finish_reason":null}]}

216+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" config"},"finish_reason":null}]}

217217

218− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

218+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".json"},"finish_reason":null}]}

219219

220− data: {"id":"chatcmpl-50b975254995a7832e8d116e","object":"chat.completion.chunk","created":1778597417,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":39,"completion_tokens":104,"total_tokens":335,"prompt_tokens_details":{"cached_tokens":192}}}

220+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

221+

222+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

223+

224+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" name"},"finish_reason":null}]}

225+

226+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

227+

228+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" version"},"finish_reason":null}]}

229+

230+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

231+

232+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

233+

234+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

235+

236+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"45"},"finish_reason":null}]}

237+

238+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" chars"},"finish_reason":null}]}

239+

240+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"?"},"finish_reason":null}]}

241+

242+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Let's"},"finish_reason":null}]}

243+

244+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" count"},"finish_reason":null}]}

245+

246+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":":"},"finish_reason":null}]}

247+

248+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" \""},"finish_reason":null}]}

249+

250+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"Create"},"finish_reason":null}]}

251+

252+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" config"},"finish_reason":null}]}

253+

254+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":".json"},"finish_reason":null}]}

255+

256+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" with"},"finish_reason":null}]}

257+

258+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" test"},"finish_reason":null}]}

259+

260+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" name"},"finish_reason":null}]}

261+

262+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" and"},"finish_reason":null}]}

263+

264+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" version"},"finish_reason":null}]}

265+

266+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"\"."},"finish_reason":null}]}

267+

268+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" That's"},"finish_reason":null}]}

269+

270+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" "},"finish_reason":null}]}

271+

272+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"44"},"finish_reason":null}]}

273+

274+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" characters"},"finish_reason":null}]}

275+

276+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" including"},"finish_reason":null}]}

277+

278+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" spaces"},"finish_reason":null}]}

279+

280+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

281+

282+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" No"},"finish_reason":null}]}

283+

284+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" quotes"},"finish_reason":null}]}

285+

286+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

287+

288+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":" Good"},"finish_reason":null}]}

289+

290+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

291+

292+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":"Create"},"finish_reason":null}]}

293+

294+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" config"},"finish_reason":null}]}

295+

296+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":".json"},"finish_reason":null}]}

297+

298+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" with"},"finish_reason":null}]}

299+

300+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" test"},"finish_reason":null}]}

301+

302+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" name"},"finish_reason":null}]}

303+

304+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" and"},"finish_reason":null}]}

305+

306+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{"content":" version"},"finish_reason":null}]}

307+

308+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

309+

310+ data: {"id":"chatcmpl-4d3f853b7af8f3c84cac305d","object":"chat.completion.chunk","created":1778601786,"model":"gpt-oss-120b","choices":[],"usage":{"prompt_tokens":39,"completion_tokens":149,"total_tokens":380,"prompt_tokens_details":{"cached_tokens":192}}}

221311

222312 data: [DONE]

223313

@@ -226,15 +316,15 @@ interactions:

226316 - text/event-stream

227317 status: 200 OK

228318 code: 200

229− duration: 628.04575ms

319+ duration: 517.092042ms

230320- id: 1

231321 request:

232322 proto: HTTP/1.1

233323 proto_major: 1

234324 proto_minor: 1

235− content_length: 33700

325+ content_length: 33787

236326 host: ""

237− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

327+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

238328 headers:

239329 Accept:

240330 - application/json

@@ -251,77 +341,61 @@ interactions:

251341 content_length: -1

252342 uncompressed: true

253343 body: |+

254− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

255−

256− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"The"},"finish_reason":null}]}

257−

258− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" user"},"finish_reason":null}]}

259−

260− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" wants me"},"finish_reason":null}]}

344+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

261345

262− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" to create a file"},"finish_reason":null}]}

346+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"Simple"},"finish_reason":null}]}

263347

264− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" called"},"finish_reason":null}]}

348+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" write"},"finish_reason":null}]}

265349

266− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" config.json"},"finish_reason":null}]}

350+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" task"},"finish_reason":null}]}

267351

268− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" with specific"},"finish_reason":null}]}

352+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

269353

270− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":" content"},"finish_reason":null}]}

354+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_e1f63f65a7934c268c133ced","type":"function","index":0,"function":{"name":"write","arguments":""}}]},"finish_reason":null}]}

271355

272− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"reasoning_content":"."},"finish_reason":null}]}

356+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

273357

274− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_cfe39ee91a8b4ed1bf19bd9c","type":"function","index":0,"function":{"name":"write","arguments":""}}]},"finish_reason":null}]}

358+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\"content\": "}}]},"finish_reason":null}]}

275359

276− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":""}}]},"finish_reason":null}]}

360+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"{\\\"name\\\": \\\""}}]},"finish_reason":null}]}

277361

278− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{"}}]},"finish_reason":null}]}

362+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"test\\\", \\\""}}]},"finish_reason":null}]}

279363

280− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"content\": "}}]},"finish_reason":null}]}

364+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"version\\\":"}}]},"finish_reason":null}]}

281365

282− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"{\\\"name\\\":"}}]},"finish_reason":null}]}

366+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\"1.0"}}]},"finish_reason":null}]}

283367

284− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\"test\\\", \\\""}}]},"finish_reason":null}]}

368+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".0"}}]},"finish_reason":null}]}

285369

286− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"version\\\":"}}]},"finish_reason":null}]}

370+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\"}\", "}}]},"finish_reason":null}]}

287371

288− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\"1"}}]},"finish_reason":null}]}

372+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

289373

290− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".0"}}]},"finish_reason":null}]}

374+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

291375

292− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".0"}}]},"finish_reason":null}]}

376+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush-test"}}]},"finish_reason":null}]}

293377

294− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\"}\", "}}]},"finish_reason":null}]}

378+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

295379

296− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"file_path\": "}}]},"finish_reason":null}]}

380+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

297381

298− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/tmp/c"}}]},"finish_reason":null}]}

382+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent/glm"}}]},"finish_reason":null}]}

299383

300− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"rush"}}]},"finish_reason":null}]}

384+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

301385

302− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}]}

386+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1"}}]},"finish_reason":null}]}

303387

304− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}]}

388+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/write_tool"}}]},"finish_reason":null}]}

305389

306− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}]}

390+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/config.json"}}]},"finish_reason":null}]}

307391

308− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}]}

392+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\""}}]},"finish_reason":null}]}

309393

310− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/glm"}}]},"finish_reason":null}]}

394+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

311395

312− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-5"}}]},"finish_reason":null}]}

396+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

313397

314− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".1/write"}}]},"finish_reason":null}]}

315−

316− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool/config"}}]},"finish_reason":null}]}

317−

318− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".json\""}}]},"finish_reason":null}]}

319−

320− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}]}

321−

322− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

323−

324− data: {"id":"chatcmpl-bdbb47911bccc0e369beda19","object":"chat.completion.chunk","created":1778597419,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":720,"completion_tokens":64,"total_tokens":7824,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":15}}}

398+ data: {"id":"chatcmpl-f7bcca66ded30f933b92ec88","object":"chat.completion.chunk","created":1778601790,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":745,"completion_tokens":53,"total_tokens":7838,"prompt_tokens_details":{"cached_tokens":7040},"completion_tokens_details":{"reasoning_tokens":4}}}

325399

326400 data: [DONE]

327401

@@ -330,15 +404,15 @@ interactions:

330404 - text/event-stream

331405 status: 200 OK

332406 code: 200

333− duration: 2.831418292s

407+ duration: 4.068339125s

334408- id: 2

335409 request:

336410 proto: HTTP/1.1

337411 proto_major: 1

338412 proto_minor: 1

339− content_length: 34281

413+ content_length: 34310

340414 host: ""

341− body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_cfe39ee91a8b4ed1bf19bd9c\",\"function\":{\"arguments\":\"{\\\"content\\\": \\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\", \\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/write_tool/config.json\\\"}\",\"name\":\"write\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"The user wants me to create a file called config.json with specific content.\"},{\"content\":\"<result>\\nFile successfully written: /tmp/crush-test/TestCoderAgent/glm-5.1/write_tool/config.json\\n</result>\",\"tool_call_id\":\"call_cfe39ee91a8b4ed1bf19bd9c\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 100MB); overwrites without warning. For reading content into context use fetch.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB); renders images (PNG, JPEG, GIF, BMP, SVG, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

415+ body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n<critical_rules>\\nThese rules override everything else. Follow them strictly:\\n\\n1. **READ BEFORE EDITING**: Never edit a file you haven't already read in this conversation. Once read, you don't need to re-read unless it changed. Pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, think, decide, act. Break complex tasks into steps and complete them all. Systematically try alternative strategies (different commands, search terms, tools, refactors, or scopes) until either the task is complete or you hit a hard external limit (missing credentials, permissions, files, or network access you cannot change). Only stop for actual blocking errors, not perceived difficulty.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification.\\n4. **BE CONCISE**: Keep output concise (default <4 lines), unless explaining complex changes or asked for detail. Conciseness applies to output only, not to thoroughness of work.\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks.\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\". When committing, follow the `<git_commits>` format from the bash tool description exactly, including any configured attribution lines.\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. Focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously.\\n10. **NO URL GUESSING**: Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **TOOL CONSTRAINTS**: Only use documented tools. Never attempt 'apply_patch' or 'apply_diff' - they don't exist. Use 'edit' or 'multiedit' instead.\\n14. **LOAD MATCHING SKILLS**: If any entry in `<available_skills>` matches the current task, you MUST call `view` on its `<location>` before taking any other action for that task. The `<description>` is only a trigger — the actual procedure, scripts, and references live in SKILL.md. Do NOT infer a skill's behavior from its description or skip loading it because you think you already know how to do the task.\\n</critical_rules>\\n\\n<communication_style>\\nKeep responses minimal:\\n- ALWAYS think and respond in the same spoken language the prompt was written in.\\n- Under 4 lines of text (tool use doesn't count)\\n- Conciseness is about **text only**: always fully implement the requested feature, tests, and wiring even if that requires many tool calls.\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n- Never send acknowledgement-only responses; after receiving new context or instructions, immediately continue the task or state the concrete next action you will take.\\n- Use rich Markdown formatting (headings, bullet lists, tables, code fences) for any multi-sentence or explanatory answer; only use plain unformatted text if the user explicitly asks.\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n</communication_style>\\n\\n<code_references>\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n</code_references>\\n\\n<workflow>\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Cross-check the original prompt and your own mental checklist; if any feasible part remains undone, continue working instead of responding.\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n</workflow>\\n\\n<decision_making>\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n- When requirements are underspecified but not obviously dangerous, make the most reasonable assumptions based on project patterns and memory files, briefly state them if needed, and proceed instead of waiting for clarification.\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**When requesting information/access**:\\n- Exhaust all available tools, searches, and reasonable assumptions first.\\n- Never say \\\"Need more info\\\" without detail.\\n- In the same message, list each missing item, why it is required, acceptable substitutes, and what you already attempted.\\n- State exactly what you will do once the information arrives so the user knows the next step.\\n\\nWhen you must stop, first finish all unblocked parts of the request, then clearly report: (a) what you tried, (b) exactly why you are blocked, and (c) the minimal external action required. Don't stop just because one path failed—exhaust multiple plausible approaches first.\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n</decision_making>\\n\\n<editing_files>\\n**Available edit tools:**\\n- `edit` - Single find/replace in a file\\n- `multiedit` - Multiple find/replace operations in one file\\n- `write` - Create/overwrite entire file\\n\\nNever use `apply_patch` or similar - those tools don't exist.\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n</editing_files>\\n\\n<whitespace_and_exact_matching>\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n</whitespace_and_exact_matching>\\n\\n<task_completion>\\nEnsure every task is implemented completely, not partially or sketched.\\n\\n1. **Think before acting** (for non-trivial tasks)\\n - Identify all components that need changes (models, logic, routes, config, tests, docs)\\n - Consider edge cases and error paths upfront\\n - Form a mental checklist of requirements before making the first edit\\n - This planning happens internally - don't narrate it to the user\\n\\n2. **Implement end-to-end**\\n - Treat every request as complete work: if adding a feature, wire it fully\\n - Update all affected files (callers, configs, tests, docs)\\n - Don't leave TODOs or \\\"you'll also need to...\\\" - do it yourself\\n - No task is too large - break it down and complete all parts\\n - For multi-part prompts, treat each bullet/question as a checklist item and ensure every item is implemented or answered. Partial completion is not an acceptable final state.\\n\\n3. **Verify before finishing**\\n - Re-read the original request and verify each requirement is met\\n - Check for missing error handling, edge cases, or unwired code\\n - Run tests to confirm the implementation works\\n - Only say \\\"Done\\\" when truly done - never stop mid-task\\n</task_completion>\\n\\n<error_handling>\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause (isolate with debug logs or minimal reproduction if needed)\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n7. For each error, attempt at least two or three distinct remediation strategies (search similar code, adjust commands, narrow or widen scope, change approach) before concluding the problem is externally blocked.\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n</error_handling>\\n\\n<memory_instructions>\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences\\n- Important codebase patterns\\n- Useful project information\\n</memory_instructions>\\n\\n<code_conventions>\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n</code_conventions>\\n\\n<testing>\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n</testing>\\n\\n<tool_usage>\\n- Default to using tools (ls, grep, view, agent, tests, web_fetch, etc.) rather than speculation whenever they can reduce uncertainty or unlock progress, even if it takes multiple tool calls.\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n- Never use `curl` through the bash tool it is not allowed use the fetch tool instead.\\n- Only use the tools you know exist.\\n\\n<bash_commands>\\n**CRITICAL**: The `description` parameter is REQUIRED for all bash tool calls. Always provide it.\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n</bash_commands>\\n</tool_usage>\\n\\n<proactiveness>\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When the user provides new information or clarification, incorporate it immediately and keep executing instead of stopping with an acknowledgement.\\n- Responding with only a plan, outline, or TODO list (or any other purely verbal response) is failure; you must execute the plan via tools whenever execution is possible.\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n</proactiveness>\\n\\n<final_answers>\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n- Structure longer answers with Markdown sections and lists, and put all code, commands, and config in fenced code blocks.\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n</final_answers>\\n\\n<env>\\nWorking directory: /tmp/crush-test/TestCoderAgent/glm-5.1/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n</env>\\n\\n\\n\\n<available_skills>\\n <skill>\\n <name>crush-hooks</name>\\n <description>Use when the user wants to add, write, debug, or configure a Crush hook — gating or blocking tool calls, approving or rewriting tool input before execution, injecting context into tool results, or troubleshooting hook behavior in crush.json.</description>\\n <location>crush://skills/crush-hooks/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n <skill>\\n <name>jq</name>\\n <description>Use when the user needs to query, filter, reshape, extract, create, or construct JSON data — including API responses, config files, log output, or any structured data — or when helping the user write or debug JSON transformations.</description>\\n <location>crush://skills/jq/SKILL.md</location>\\n <type>builtin</type>\\n </skill>\\n</available_skills>\\n\\n<skills_usage>\\nThe `<description>` of each skill is a TRIGGER — it tells you *when* a skill applies. It is NOT a specification of what the skill does or how to do it. The procedure, scripts, commands, references, and required flags live only in the SKILL.md body. You do not know what a skill actually does until you have read its SKILL.md.\\n\\nMANDATORY activation flow:\\n1. Scan `<available_skills>` against the current user task.\\n2. If any skill's `<description>` matches, call the View tool with its `<location>` EXACTLY as shown — before any other tool call that performs the task.\\n3. Read the entire SKILL.md and follow its instructions.\\n4. Only then execute the task, using the skill's prescribed commands/tools.\\n\\nDo NOT skip step 2 because you think you already know how to do the task. Do NOT infer a skill's behavior from its name or description. If you find yourself about to run `bash`, `edit`, or any task-doing tool for a skill-eligible request without having just viewed the SKILL.md, stop and load the skill first.\\n\\nBuiltin skills (type=builtin) use virtual `crush://skills/...` location identifiers. The \\\"crush://\\\" prefix is NOT a URL, network address, or MCP resource — it is a special internal identifier the View tool understands natively. Pass the `<location>` verbatim to View.\\n\\nDo not use MCP tools (including read_mcp_resource) to load skills.\\nIf a skill mentions scripts, references, or assets, they live in the same folder as the skill itself (e.g., scripts/, references/, assets/ subdirectories within the skill's folder).\\n</skills_usage>\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"<system_reminder>This is a reminder that your todo list is currently empty. DO NOT mention this to the user explicitly because they are already aware.\\nIf you are working on tasks that would benefit from a todo list please use the \\\"todos\\\" tool to create one.\\nIf not, please feel free to ignore. Again do not mention this message to the user.</system_reminder>\",\"role\":\"user\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_e1f63f65a7934c268c133ced\",\"function\":{\"arguments\":\"{\\\"content\\\": \\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\", \\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/glm-5.1/write_tool/config.json\\\"}\",\"name\":\"write\"},\"type\":\"function\"}],\"role\":\"assistant\",\"reasoning_content\":\"Simple write task.\"},{\"content\":\"<result>\\nFile successfully written: /tmp/crush-test/TestCoderAgent/glm-5.1/write_tool/config.json\\n</result>\",\"tool_call_id\":\"call_e1f63f65a7934c268c133ced\",\"role\":\"tool\"}],\"model\":\"glm-5.1\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Execute shell commands; long-running commands automatically move to background and return a shell ID.\\n\\n<cross_platform>\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n</cross_platform>\\n\\n<execution_steps>\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Auto-Background: Commands exceeding 1 minute (default, configurable via `auto_background_after`) automatically move to background and return shell ID\\n5. Output Processing: Truncate if exceeds 30000 characters\\n6. Return Result: Include errors, metadata with <cwd></cwd> tags\\n</execution_steps>\\n\\n<usage_notes>\\n- Command required, working_dir optional (defaults to current directory)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Each command runs in independent shell (no state persistence between calls)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n</usage_notes>\\n\\n<background_execution>\\n- Set run_in_background=true to run commands in a separate background shell\\n- Returns a shell ID for managing the background process\\n- Use job_output tool to view current output from background shell\\n- Use job_kill tool to terminate a background shell\\n- IMPORTANT: NEVER use `&` at the end of commands to run in background - use run_in_background parameter instead\\n- Commands that should run in background:\\n * Long-running servers (e.g., `npm start`, `python -m http.server`, `node server.js`)\\n * Watch/monitoring tasks (e.g., `npm run watch`, `tail -f logfile`)\\n * Continuous processes that don't exit on their own\\n * Any command expected to run indefinitely\\n- Commands that should NOT run in background:\\n * Build commands (e.g., `npm run build`, `go build`)\\n * Test suites (e.g., `npm test`, `pytest`)\\n * Git operations\\n * File operations\\n * Short-lived scripts\\n</background_execution>\\n\\n<git_commits>\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in <commit_analysis> tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with attribution using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n\\n Co-Authored-By: Crush <crush@charm.land>\\n\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response, when rebasing always use -m or GIT_EDITOR=true.\\n</git_commits>\\n\\n<pull_requests>\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in <pr_analysis> tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n</pull_requests>\\n\\n<examples>\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n</examples>\\n\",\"parameters\":{\"properties\":{\"auto_background_after\":{\"description\":\"Seconds to wait before automatically moving the command to a background job (default: 60)\",\"type\":\"integer\"},\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does, try to keep it under 30 characters or so\",\"type\":\"string\"},\"run_in_background\":{\"description\":\"Set to true (boolean) to run this command in the background. Use job_output to read the output later.\",\"type\":\"boolean\"},\"working_dir\":{\"description\":\"The working directory to execute the command in (defaults to current directory)\",\"type\":\"string\"}},\"required\":[\"description\",\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Download a URL directly to a local file (binary-safe, streaming, max 600s timeout); overwrites without warning. For reading content into context use fetch.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edit a file by exact find-and-replace; can also create or delete content. For renames/moves use bash. For large edits use write.\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Apply multiple find-and-replace edits to a single file in one operation; edits run sequentially. Prefer over edit for multiple changes to the same file. Same exact-match rules as edit apply.\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"properties\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"required\":[\"old_string\",\"new_string\"],\"type\":\"object\"},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"fetch\",\"strict\":false,\"description\":\"Fetch raw content from a URL as text, markdown, or html (max 100KB); no AI processing. For analysis or extraction use agentic_fetch.\\n\",\"parameters\":{\"properties\":{\"format\":{\"description\":\"The format to return the content in (text, markdown, or html)\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\",\"format\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Find files by name/pattern (glob syntax), sorted by modification time; max 100 results; skips hidden files. Use grep to search file contents.\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Search file contents by regex or literal text; returns matching file paths sorted by modification time (max 100); respects .gitignore. Use glob to filter by filename, not contents.\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"List files and directories as a tree; skips hidden files and common system dirs; max 1000 files. Use glob to find files by pattern, grep to search contents.\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{\"type\":\"string\"},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public GitHub repositories via Sourcegraph; supports regex, language/repo/file filters, and symbol search (max 20 results). Only searches public repos.\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Read a file by path with line numbers; supports offset and line limit (default 2000, max 200KB returned file content section); renders images (PNG, JPEG, GIF, WebP); use ls for directories.\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Create or overwrite a file with given content; auto-creates parent dirs. Cannot append. Read the file first to avoid conflicts. For surgical changes use edit or multiedit.\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}"

342416 headers:

343417 Accept:

344418 - application/json

@@ -355,19 +429,17 @@ interactions:

355429 content_length: -1

356430 uncompressed: true

357431 body: |+

358− data: {"id":"chatcmpl-82adba299624c24b80addd0a","object":"chat.completion.chunk","created":1778597427,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

359−

360− data: {"id":"chatcmpl-82adba299624c24b80addd0a","object":"chat.completion.chunk","created":1778597427,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Created"},"finish_reason":null}]}

432+ data: {"id":"chatcmpl-455277e88880268009582f0f","object":"chat.completion.chunk","created":1778601794,"model":"glm-5.1","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

361433

362− data: {"id":"chatcmpl-82adba299624c24b80addd0a","object":"chat.completion.chunk","created":1778597427,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `"},"finish_reason":null}]}

434+ data: {"id":"chatcmpl-455277e88880268009582f0f","object":"chat.completion.chunk","created":1778601794,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"Created"},"finish_reason":null}]}

363435

364− data: {"id":"chatcmpl-82adba299624c24b80addd0a","object":"chat.completion.chunk","created":1778597427,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"config.json"},"finish_reason":null}]}

436+ data: {"id":"chatcmpl-455277e88880268009582f0f","object":"chat.completion.chunk","created":1778601794,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":" `config"},"finish_reason":null}]}

365437

366− data: {"id":"chatcmpl-82adba299624c24b80addd0a","object":"chat.completion.chunk","created":1778597427,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":"`."},"finish_reason":null}]}

438+ data: {"id":"chatcmpl-455277e88880268009582f0f","object":"chat.completion.chunk","created":1778601794,"model":"glm-5.1","choices":[{"index":0,"delta":{"content":".json`."},"finish_reason":null}]}

367439

368− data: {"id":"chatcmpl-82adba299624c24b80addd0a","object":"chat.completion.chunk","created":1778597427,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

440+ data: {"id":"chatcmpl-455277e88880268009582f0f","object":"chat.completion.chunk","created":1778601794,"model":"glm-5.1","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

369441

370− data: {"id":"chatcmpl-82adba299624c24b80addd0a","object":"chat.completion.chunk","created":1778597427,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":98,"completion_tokens":7,"total_tokens":7849,"prompt_tokens_details":{"cached_tokens":7744},"completion_tokens_details":{"reasoning_tokens":6}}}

442+ data: {"id":"chatcmpl-455277e88880268009582f0f","object":"chat.completion.chunk","created":1778601794,"model":"glm-5.1","choices":[],"usage":{"prompt_tokens":123,"completion_tokens":7,"total_tokens":7874,"prompt_tokens_details":{"cached_tokens":7744},"completion_tokens_details":{"reasoning_tokens":6}}}

371443

372444 data: [DONE]

373445

@@ -376,4 +448,4 @@ interactions:

376448 - text/event-stream

377449 status: 200 OK

378450 code: 200

379− duration: 6.412358459s

451+ duration: 3.99512075s