claude-hooks-developer

Hooks intercept tool calls at specific lifecycle events like PreToolUse and PostToolUse, letting you validate operations before execution. Create a hook script in .claude/hooks, configure it in .claude/settings.json with event matchers, and use exit codes to approve or block actions.

PreToolUse hooks run before a tool executes, letting you validate inputs and block unsafe commands. PostToolUse hooks run after execution to log results or trigger notifications. Use PreToolUse for security gates and PostToolUse for auditing and downstream actions.

File protection and command validation use hook scripts to inspect tool inputs against blocklists or rules. Extract tool parameters, match against patterns, and return decision objects with exit codes 0 (allow) or 1 (deny) to enforce security policies.

Yes. Hooks can access CLAUDE_PROJECT_DIR and tool input data to make context-aware decisions. Use these variables in Python or Bash scripts to reference project structure and dynamically apply validation rules.

Beyond PreToolUse and PostToolUse, available events include UserPromptSubmit, Stop, SessionStart, Notification, and PermissionRequest. Each event lets you intercept different workflow stages for logging, validation, or notification routing.

Make hook scripts executable in .claude/hooks and add configuration to .claude/settings.json. No special permissions are required; test by triggering relevant tool operations and verifying exit codes and decision objects match your intent.