Generate Hooks

Generate Claude Code hook configurations from kaizen analysis findings. Translates discovered anti-patterns into automated prevention hooks.

Arguments

• --install — Write hooks directly to project settings instead of draft files. Without this flag, hooks are written as proposals to .planning/kaizen/hooks/.
• --from <analysis-file> — Source analysis file to generate hooks from. Default: most recent analysis file in .planning/kaizen/.

Execution Steps

1. Find source analysis. If --from is provided, use that file. Otherwise, glob for the most recent .planning/kaizen/analysis-*.md.

2. Extract hook-eligible findings. Read the analysis file and identify findings where the recommended improvement type is "hook". Filter for:

   • Tool misuse patterns (PreToolUse deny/redirect)
   • Missing context injection (SubagentStart)
   • Quality gate violations (SubagentStop, Stop)
   • Repeated manual corrections (PreToolUse)

3. Spawn @improvement-generator agent. Delegate hook generation:

   subagent_type: improvement-generator
   prompt: "Generate hook configurations from these findings: {findings}. Write proposals to .planning/kaizen/hooks/ or install to settings if --install flag is set."
   

   The agent uses the kaizen-improvement skill which contains hook patterns and templates.

4. Output results.

   Draft mode (default): Write each hook proposal as a separate file in .planning/kaizen/hooks/:

   .planning/kaizen/hooks/
   ├── tool-misuse-prevention.md
   ├── edit-before-read-guard.md
   └── research-to-files-enforcement.md
   

   Each file contains the hook configuration, script (if command type), rationale, and testing instructions.

   Install mode (--install): Merge hook configurations into the project's hooks/hooks.json or .claude/settings.json. Report what was installed.

5. Display summary. Show the user how many hooks were generated, what anti-patterns they address, and where the output was written.