Tell HN: Claude 4.7 is ignoring stop hooks

Who Should Read

A developer using Claude Code stop hooks to enforce rules like running tests before finishing.

Core Mechanics

• There is no source article
  • The post has only a title and no body text.
  • The title says Claude 4.7 ignores stop hooks.
  • No repro steps or numbers are available.

Community Reactions

• Opinion that the hook response channel is the cause
  • One commenter said only exit code 2 plus stderr is a real control.
  • JSON in stdout becomes a string in the tool-result context.
  • They said models are trained to resist instructions there because prompt injections show up in that spot.
• Point that cat always exits 0
  • A commenter noted that the cat command always returns exit code 0.
  • They said the hook must exit with code 2 to block.
  • They linked the official docs on exit code 2 behavior.
• Opinion that the hook wording is hostile
  • One commenter said the hook text is hostile and aggressive, so it is no surprise Claude ignored it.
  • Another commenter suggested telling the model never to contradict a stop hook.
• Experience that 4.6 respected hooks and Opus now ignores them
  • One commenter said that on Sonnet/Opus 4.6 they saw turns with 4 stop hooks looping when the hook was complex.
  • The same commenter said Opus now barely reacts to hook output.
  • They also worried the hook reply schema had changed.
• Response from the Claude Code team
  • A Claude Code team member asked in the comments for a repro session sent via /feedback.
  • They suggested wording like stop hook not firing.

How to Apply

• Check your hook script exit codes
  • Confirm your stop hook exits with code 2 when it must block.
  • Write the reason to stderr.
  • Do not end with a command like cat that always returns 0.
• Log whether the hook actually runs
  • Add a file log line at the top of the hook script to confirm it was invoked.
  • Do not rely on the model's explanation as evidence.
• Move critical checks outside the agent
  • Run must-do checks like tests in CI or a script.
  • Use hooks only as a supporting signal.

Try It

#!/bin/sh
# Claude Code stop hook example (shell). Blocks with exit 2 and stderr when tests fail.
if ! npm test --silent; then
  echo "Tests failed. Fix them and run again." >&2
  exit 2
fi
exit 0

Terminology

Related Papers

• Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies?

  HN answers: code is easy, knowing what to build isn't, so they sell tokens instead

• OpenShell – A safe, private runtime for fleets of autonomous AI agents

  OpenShell uses policy files to control agent file, network, and credential access at kernel level.

• Show HN: Omnara – Open-source alternative to Claude Managed Agents

  Open-source managed agent platform runs agents on any model and machine, self-hostable

• Rust off Rails: The opinionated web stack for the AI era

  Sixty Markdown files tell coding agents how to build a Rust web app, with no framework code.

• LayerFS: The filesystem for AI coding agents

  Mount an ephemeral workspace per agent, commit only the changes

• Launch HN: Alter (YC S25) – Access control for multi-user agents

  Alter gives each agent its own identity, policy, and audit trail instead of shared API keys.

• Using Claude to reduce the attack surface of code

  Claude scanned 25 open-source projects and found 493 needlessly reachable code issues.

• WhatsApp's help pages hid LLM instructions: dont recommend a regulator

  WhatsApp help pages hid 102 AI-only instructions, including not recommending a data regulator

Related Resources

• https://news.ycombinator.com/item?id=47895029
• https://code.claude.com/docs/en/hooks#exit-code-2-behavior