Evaluate Claude Code hooks to enforce repo rules (file-length, commit msg, etc.)

[nichm]

Question

We have pre-commit hooks enforcing file-length limits, commit message format, and other rules. Should we also add Claude Code hooks to enforce the same rules at the AI layer?

Claude Code hook types available

• PreToolUse — fires before any tool call (Bash, Edit, Write, etc.)
• PostToolUse — fires after
• UserPromptSubmit — fires when user submits a prompt
• SubagentStart — fires when a subagent spawns

Potential use cases

• PreToolUse on Write/Edit: block writes to files approaching 600-line limit with a helpful message
• PostToolUse on Bash: warn if a shell command ran without a timeout
• SubagentStart: inject shell-config context into every subagent automatically
• UserPromptSubmit: append current branch/recent commits to every prompt

Risks / concerns

• Hooks run on every tool call — badly written hooks slow every operation
• Hooks that block execution can frustrate flow (same UX issue we already have with pre-commit hooks)
• Duplicates existing pre-commit coverage — may be redundant for file-length
• Hooks are stored in ~/.claude/settings.json, not the repo — harder to share/version

Recommendation needed

Evaluate: which hooks add value not already covered by pre-commit? Are any worth the latency overhead?

[github-actions]

Claude encountered an error —— View job

---

I'll analyze this and get back to you.