What is a Claude Code hook?

A hook is a specialized script or LLM prompt that executes at specific points in the Claude Code workflow, allowing you to intercept, validate, or modify the AI's actions.

Where are hook settings stored?

Hooks are configured in settings.json files located in your ~/.claude/ directory for global rules or the .claude/ directory within your project for local rules.

Can I use hooks to auto-format code?

Yes, by setting up a PostToolUse hook on 'Write' or 'Edit' events, you can automatically trigger formatters like Prettier, ESLint, or Black after Claude modifies a file.

What languages can I use to write hooks?

Hooks can be any executable script, including Bash, Python, or Node.js. They communicate with Claude Code using JSON via standard input (stdin) and output (stdout).

How do I block dangerous commands with hooks?

You can use a PreToolUse hook with a Bash matcher. If the script detects a dangerous command, it can return an exit code of 2 to block the action and notify the AI.

Claude Code Hooks Automator

Name: Claude Code Hooks Automator
Author: hgeldenhuys

byhgeldenhuys

•

개발자 도구

Implements event-driven automation for Claude Code to handle validation, auto-formatting, and custom workflow triggers.

The Creating Hooks skill provides a comprehensive framework for building and managing custom scripts that intercept Claude Code's lifecycle events. By leveraging hooks, developers can automate repetitive tasks like code linting, enforce security policies by blocking dangerous commands, and inject project-specific context during session starts. This skill bridges the gap between AI-driven code generation and professional development standards by ensuring that every tool execution—from bash commands to file edits—adheres to your project's specific requirements and safety guidelines.

주요 기능

01Security-focused command guarding to block or validate dangerous terminal operations

02Comprehensive event triggers including PreToolUse, PostToolUse, and SessionStart

03Support for both Bash-based command hooks and LLM-powered prompt hooks

041 GitHub stars

05YAML-based Hook Framework for simplified management of complex automation chains

06Automated code formatting and linting integration for Write and Edit tools

사용 사례

01Preventing accidental execution of destructive shell commands using regex-based matchers

02Automatically running Prettier or Black after Claude modifies a file to ensure style consistency

03Injecting real-time environment data like Git branch info and database status into the session context

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add hgeldenhuys/claude-code-sdk creating-hooks

For use in Claude.ai and ChatGPT

Download Skill