Can I use hooks to block dangerous commands?

Yes, by returning an exit code of 2 in your hook script, you can signal Claude to block the action and provide feedback to the LLM about why it was stopped.

What is a Claude Code hook?

A hook is a custom shell script or prompt that Claude Code executes automatically when specific events occur, such as before or after a tool is used or when a session starts.

Where are hook configurations and scripts stored?

Hook settings are defined in .claude/settings.json, and the corresponding executable scripts are typically stored in the .claude/hooks/ directory.

Why aren't my hook changes taking effect immediately?

Hooks are snapshotted at startup for security purposes. You must restart Claude Code or run the /hooks command to review and apply the updated configuration.

Claude Code Hook Development

Name: Claude Code Hook Development
Author: dwmkerr

bydwmkerr

•

Productivity & Workflow

Automates workflows and enforces guardrails by creating custom event-driven shell hooks for the Claude Code CLI.

This skill empowers developers to extend Claude Code's functionality by implementing event-driven hooks that trigger shell commands or LLM prompts. By integrating with core lifecycle events like tool usage, permission requests, or session starts, users can build robust guardrails, automate repetitive tasks such as linting and testing, and enforce organizational policies directly within their AI-driven development environment. It provides a standardized framework for managing these extensions, including security-focused snapshots, wildcard tool matching, and detailed input/output schemas for sophisticated automation logic.

Key Features

01Flexible wildcard matching for tools like Bash, npm, or Git commands

02Event-driven shell command execution for Pre and Post-tool usage

03Context-aware Prompt hooks for intelligent, AI-driven decision making

042 GitHub stars

05Custom guardrails to block dangerous operations using specific exit code patterns

06Comprehensive lifecycle event support including session management and subagent tasks

Use Cases

01Automatically running linters and formatters after Claude edits source code files

02Creating a security firewall to block or audit specific high-risk CLI commands

03Preventing accidental pushes to protected Git branches like main or master

Key Features

01Flexible wildcard matching for tools like Bash, npm, or Git commands

02Event-driven shell command execution for Pre and Post-tool usage

03Context-aware Prompt hooks for intelligent, AI-driven decision making

042 GitHub stars

05Custom guardrails to block dangerous operations using specific exit code patterns

06Comprehensive lifecycle event support including session management and subagent tasks

Use Cases

01Automatically running linters and formatters after Claude edits source code files

02Creating a security firewall to block or audit specific high-risk CLI commands

03Preventing accidental pushes to protected Git branches like main or master