claude-code - 💡(How to fix) Fix PreToolUse/PostToolUse hooks always show 'hook error' on Windows despite exit 0 [1 comments, 2 participants]

On Windows 11 with Claude Code 2.1.96, all PreToolUse and PostToolUse hooks display "hook error" in the transcript even when they exit with code 0 and produce correct output. SessionStart and UserPromptSubmit hooks display "success" correctly. 5. Observe: PreToolUse:Bash hook error appears in transcript | node -e "process.exit(0)" | 0 | none | none | error | | bash -c "exit 0" | 0 | none | none | error | | bash -c "cat; exit 0" (stdin passthrough) | 0 | JSON | none | error | | Node script outputting {} | 0 | {} | none | error | | Node script with stderr + exit 0 | 0 | JSON | text | error | All hooks execute correctly (verified via log files and side effects). The "hook error" label is purely a display issue. Every tool use (Bash, Read, Edit, Write) generates 8-14 "hook error" lines in the transcript, making it very noisy and hard to distinguish real errors from cosmetic ones.

TL;DR

The issue can be mitigated by modifying the hook configuration to handle the output and error streams correctly.

Guidance

• Verify that the hook commands are executing correctly and producing the expected output by checking the log files and side effects.
• Check the Claude Code documentation for any specific requirements or guidelines for handling output and error streams in hooks.
• Consider modifying the hook commands to explicitly handle and redirect output and error streams, for example, by appending > /dev/null 2>&1 to the command.
• Test the hooks with different commands and output scenarios to ensure the issue is not specific to a particular command or output type.

Example

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [
          {
            "type": "command",
            "command": "node -e \"process.exit(0)\" > /dev/null 2>&1"
          }
        ]
      }
    ]
  }
}

Notes

The issue seems to be related to how Claude Code handles the output and error streams of the hook commands. The provided examples suggest that the issue is not specific to a particular command or output type, but rather a general issue with how the hooks are executed and displayed in the transcript.

Recommendation

Apply workaround: Modify the hook configuration to handle the output and error streams correctly, as the root cause of the issue is likely related to how Claude Code handles these streams.