I kept writing the same hooks for different AI coding tools with slightly different configs:

- Claude Code: ~/.claude/settings.json (PreToolUse)

- Cursor: ~/.cursor/hooks.json (beforeShellExecution)

- Gemini CLI: ~/.gemini/settings.json (BeforeTool)

- OpenCode: ES module plugins

So I made a skill that unifies them. One hook script works across all 4 tools.

It also includes patterns for common use cases: blocking dangerous commands, auto-formatting, notifications, audit logging.

Built two apps using Claude Code + this skill:

- Code Buddy (SwiftUI): menu bar status for AI sessions

- Veto (Rust): risk scoring + Touch ID before dangerous commands

Source: https://github.com/runkids/ai-hooks-integration

Feedback welcome.