| name | update-shared-hooks |
| description | Add, migrate, or update shared hook scripts and wire them into Codex, Claude Code, and OpenCode with tool-specific hook wrappers. Use when the user asks to share hooks, install pre/post tool-use automation, change hook scripts in the shared agents folder, audit hook configuration, or preserve one hook implementation across all three agent environments. |
Update Shared Hooks
Overview
Use this skill to keep hook logic in one shared script while configuring each agent tool with its own supported hook mechanism. Shared implementation belongs in ~/.config/agents/hooks.
Workflow
- Read
~/.config/AGENTS.md and ~/.config/agents/AGENTS.md.
- Inspect existing hook config files and plugin folders before editing.
- Put reusable hook implementation in
~/.config/agents/hooks/<script>.
- Add only thin tool-specific wrappers or config entries that call the shared script.
- Preserve existing hooks and plugin behavior. Merge instead of replacing.
- Keep secrets and machine-local paths out of shared hook scripts unless the user explicitly accepts local-only behavior.
- Validate syntax for edited config files and run the hook script directly with safe sample input when possible.
- Capture rendered-dotfile changes with
chezmoi add.
Guardrails
- Do not symlink one universal hook config across all tools. Hook config formats differ.
- Do not overwrite existing
settings.json, config.toml, hooks.json, opencode.json, or plugin files without inspecting them.
- Make shared hook scripts executable only when required, and preserve file mode through chezmoi if needed.
- Prefer deterministic scripts with clear stdin/stdout behavior. Avoid interactive prompts inside hooks.
- If a hook can block agent operation, add failure behavior deliberately and document it.
Tool Map
Read references/hook-tool-map.md before wiring a shared script into active hook config.
Output Standard
Report the shared script path, every active hook wrapper changed, how existing hooks were preserved, and what validation was run.
