claude-code - 💡(How to fix) Fix [Bug] PostToolUse "block" reason is additive not substitutive; SubagentStop exit-2 behavior undocumentedroceed [1 participants]

TL;DR

The issue may be resolved by updating the documentation for PostToolUse and SubagentStop hooks to reflect their actual behavior, particularly regarding the handling of tool_response.output and reason.

Guidance

• Review the current implementation of PostToolUse and SubagentStop hooks to understand how they handle tool_response.output and reason.
• Verify that the documentation accurately reflects the empirical behavior observed, including the appending of reason to the parent agent's transcript.
• Consider updating the documentation to clarify that the original tool_response.output is not prevented from being seen by Claude, but rather is included in the parent agent's transcript.
• Test the updated documentation with different scenarios to ensure it accurately reflects the behavior of the hooks.

Notes

The provided issue description is truncated, so the full scope of the problem and potential solutions may not be entirely clear. Further investigation into the specific behaviors of PostToolUse and SubagentStop hooks may be necessary to fully address the issue.

Recommendation

Apply workaround: Update the documentation to reflect the empirical behavior of the PostToolUse and SubagentStop hooks, as this will at least clarify the current functionality for users, even if the underlying behavior is not changed.