create-hooks

Event-driven hooks automate actions triggered by Claude Code events like PreToolUse, PostToolUse, SessionStart, and UserPromptSubmit. Define hook configurations specifying event types, matcher patterns, and actions—such as validating commands, blocking unsafe operations, or injecting context—without modifying core code.

Claude Code supports command hooks and prompt hooks. Command hooks intercept and validate tool executions; prompt hooks process user input. Both types use matcher rules, input/output schemas, and permission controls to log activity, modify inputs, block unsafe actions, and enforce custom completion criteria.

Use the `claude --debug` flag to test hooks and verify behavior. Debug mode outputs hook execution details, matcher rule evaluation, and input/output transformations, helping you identify configuration errors and validate that hooks trigger on the correct events and modify data as intended.

Yes. PreToolUse hooks intercept tool inputs before execution, allowing you to validate commands, block unsafe operations, modify parameters, and inject context. PostToolUse hooks process tool outputs for logging, filtering, or transformation after execution completes.

Hook configuration requires specifying the event type (PreToolUse, PostToolUse, Stop, SessionStart, etc.), matcher patterns to filter which operations trigger the hook, hook type (command or prompt), input/output schemas, environment variables, and permission controls defining allowed actions and safety checks.

Hooks enable workflow automation, validation, and logging without touching core code. They provide a declarative, event-driven layer for implementing custom criteria, access controls, activity tracking, and context injection—keeping your automation logic separate and maintainable.