moai-cc-hooks

Safety hooks in Claude Code use PreToolUse hooks to validate commands before execution. Configure them in settings.json with matcher-based routing, define validation logic, and return exit code 0 to allow or 1 to block execution. PreToolUse hooks execute in under 100ms to avoid workflow delays.

Claude Code supports five hook types: PreToolUse (validate before execution), PostToolUse (process results), SessionStart (initialize context), Notification (alert on events), and Stop (cleanup on exit). Each hook type enables different automation scenarios from safety gates to context seeding.

Yes. PostToolUse and SessionStart hooks can trigger auto-formatting and linting workflows without blocking work. Configure command execution guards in settings.json and use matcher-based routing to apply formatting and linting rules selectively across your session.

Claude Code hooks use JSON-based settings.json configuration with matcher-based routing to direct hook behavior. Define matchers to target specific commands or contexts, assign hook handlers, and specify exit codes (0 for success, 1 for block, 2 for error). Logging guidelines ensure visibility into hook execution.

PreToolUse hooks must execute in under 100ms to maintain responsiveness. SessionStart hooks have a 500ms limit. These constraints ensure hooks enforce safety and automation without slowing down Claude Code workflows or creating noticeable delays.

Claude Code hooks have no external dependencies. Setup requires only JSON-based settings.json configuration and script components. Exit code semantics (0/1/2) and comprehensive logging provide robust hook behavior without additional tools or prerequisites.