julien-dev-hook-creator

Claude Code hooks are deterministic shell commands or prompts executed at specific lifecycle events like SessionStart, SessionEnd, PreToolUse, PostToolUse, and UserPromptSubmit. Define hooks in ~/.claude/settings.json or .claude/settings.json with a command block and timeout to enable validation, logging, environment setup, and external-tool integration.

Hooks can be configured in two locations: ~/.claude/settings.json for global hooks across all Claude Code sessions, or .claude/settings.json in a project directory for project-specific hooks. Both support the same hook template structure and settings.

Claude Code hooks use exit code 0 to continue execution and exit code 2 to block or prevent the lifecycle event from proceeding. These deterministic exit codes enable hooks to enforce validation and gating logic across automation workflows.

Yes. Hooks run deterministically at defined lifecycle events, supporting validation, logging, and environment setup across both global and project configurations. Input and output are handled as JSON, making hooks compatible with external-tool integration and cross-session automation.

Claude Code hooks support SessionStart, SessionEnd, PreToolUse, PostToolUse, and UserPromptSubmit events. Each event enables different automation patterns—SessionStart for environment setup, PreToolUse for validation, PostToolUse for logging, and UserPromptSubmit for prompt-level gating.

Hooks can be defined as either deterministic shell commands or LLM prompts, so you can use shell scripting, structured configuration, or prompt-based logic depending on your automation need. Shell command hooks must return defined exit codes (0 or 2) and handle JSON input/output.