What are Claude Code hooks?

Hooks are user-defined shell commands or HTTP requests that trigger at specific points in Claude's lifecycle, allowing you to observe, modify, or block its behavior.

How do I configure a new hook?

Hooks are configured in your project or user settings.json file by defining the event type, a matcher for specific tools, and the command or script to execute.

What format do hooks use for communication?

Hooks receive state information via stdin as structured JSON and must return their decisions or modifications via stdout, also in JSON format.

Can hooks block Claude from running commands?

Yes, hooks like PreToolUse and UserPromptSubmit can block actions by returning a 'deny' decision or using exit code 2, which is useful for security enforcement.

Do hooks slow down Claude Code's performance?

Because hooks run synchronously, Claude waits for them to complete. It is best practice to keep hook logic fast and use specific matchers to avoid unnecessary execution.

Claude Code Hooks System

Name: Claude Code Hooks System
Author: markus41

bymarkus41

•

Productivity & Workflow

Extends Claude Code functionality by executing custom logic and validations at key points in the development lifecycle.

The Claude Code Hooks System provides a robust framework for intercepting and influencing Claude's behavior through user-defined shell commands, HTTP requests, and sub-agent evaluations. By tapping into lifecycle events such as tool execution, session starts, and task completion, developers can implement custom security guardrails, automate activity logging, and enforce project-specific quality standards. This skill serves as the definitive reference for configuring matchers and managing JSON-based input/output to create a tailored, secure, and highly automated AI-driven development environment.

Key Features

01Full lifecycle event interception including PreToolUse and PostToolUse

02Advanced matcher patterns for granular control over specific tools

03Blocking mechanisms to enforce security policies and prevent dangerous actions

049 GitHub stars

05Support for command, http, prompt, and agent hook types

06Structured JSON communication for modifying tool inputs and outputs

Use Cases

01Automating project-wide logging and error capturing for local 'lessons learned' documentation

02Implementing security guardrails to block destructive shell commands or unauthorized access

03Triggering desktop or Slack notifications when long-running AI tasks are completed

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add markus41/claude hooks-system

For use in Claude.ai and ChatGPT