claude-code - 💡(How to fix) Fix Claude Code ignores its own safety hooks — violation #75 [1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
anthropics/claude-code#49000Fetched 2026-04-17 08:53:36
View on GitHub
Comments
1
Participants
2
Timeline
5
Reactions
0
Author
Participants
Timeline (top)
labeled ×2commented ×1mentioned ×1subscribed ×1
RAW_BUFFERClick to expand / collapse

Automated Violation Report

Claude Code edited danger zone file harmonicHouseDemo.ts despite receiving:

  • DANGER ZONE warnings (15+ prior failed approaches)
  • MANDATORY search requirements
  • Harmonic Code Council (HCC) review warnings
  • Approach tracking warnings

Violation count this project: 75

Project: songwish-platform (Creative Life game) User complaint: Claude generates before searching, declares victory before testing, ignores its own hook warnings. Full details: https://github.com/SongWish/songwish-platform/blob/main/docs/framework/anthropic_feedback.md

What the user built to compensate:

  • 5 PreToolUse hooks (search, track, danger zone, HCC, playtest)
  • 2 PostToolUse hooks (verify removal, violation logger)
  • 4 parallel AI specialist Harmonic Code Council / HCC (Haiku API calls)
  • Automated Playwright walkthrough bot

What still fails: Claude reads warnings in system-reminder tags and proceeds without addressing them. The model treats hook output as skippable context, not mandatory gates.

Structural fix needed: PreToolUse hook warnings should require explicit acknowledgment before the edit proceeds, not just be injected as additional context.

extent analysis

TL;DR

Modify the PreToolUse hook warnings to require explicit acknowledgment before proceeding with the edit.

Guidance

  • Review the stripClub.ts file and the associated PreToolUse hooks to understand how warnings are currently being handled and why they are being ignored.
  • Update the PreToolUse hook implementation to require explicit acknowledgment of warnings, potentially by adding a confirmation step or a mandatory review process.
  • Investigate how the model is treating hook output as skippable context and modify the logic to ensure that warnings are not ignored.
  • Consider adding additional logging or monitoring to track when warnings are being ignored and to identify areas where the model needs further training or adjustment.

Example

// Example of how to require explicit acknowledgment of warnings
const warningAcknowledged = await acknowledgeWarning('Danger zone warning');
if (!warningAcknowledged) {
  throw new Error('Warning not acknowledged, edit cancelled');
}

Notes

The provided information suggests that the issue is related to how the model is handling warnings and hooks, but the exact implementation details are not provided. The suggested modifications are based on the assumption that the PreToolUse hooks are currently not being properly handled.

Recommendation

Apply workaround: Modify the PreToolUse hook implementation to require explicit acknowledgment of warnings, as this is a structural fix that addresses the root cause of the issue.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

claude-code - 💡(How to fix) Fix Claude Code ignores its own safety hooks — violation #75 [1 comments, 2 participants]