claude-code-hooks

PreToolUse hooks run custom Bash scripts that validate inputs before tool execution. Exit code 0 allows the tool to proceed, while exit code 2 blocks execution. You define matcher patterns to apply hooks selectively per tool, receiving JSON input via stdin with full context about the tool call.

Yes. PreToolUse hooks enforce permission workflows by inspecting tool calls and blocking them with exit code 2 if they violate your policies. Hooks receive environment variables with project, session, and tool context, letting you make informed block/allow decisions before execution.

PostToolUse hooks run after tool execution completes, enabling audit logging and metrics collection. They receive tool results via JSON stdin and can emit notifications or write audit records. This approach captures what executed, when, and with what outcome without blocking workflows.

Matcher-based event selection lets you target hooks to particular tools by name. Configure PreToolUse or PostToolUse hooks with matcher patterns so they run only for relevant tools like Bash or file operations, reducing noise and overhead.

Hooks support timeout handling to prevent validators from stalling Claude Code workflows. Define appropriate timeouts for your Bash hook scripts; timeouts are treated as non-blocking failures, allowing execution to proceed while you log the issue.

PostToolUse hooks enable notification emission by running scripts that detect tool patterns and trigger alerts. Combined with environment variables providing context, hooks can route notifications to logging systems, chat platforms, or monitoring tools after operations complete.