Can I prevent Claude from running specific commands?

Yes, by setting the action field to 'block' in the rule's frontmatter, you can prevent the execution of matched bash commands or stop the session during specific events.

Do I need to restart Claude Code after adding a rule?

No, Hookify rules are read dynamically. Any changes or new files in the .claude/ directory take effect on the very next tool use.

Where should Hookify rules be stored?

Rules should be stored in the .claude/ directory of your project using the naming convention hookify.{rule-name}.local.md.

What is a Hookify rule?

A Hookify rule is a markdown file with YAML frontmatter that defines patterns Claude should watch for. When a pattern matches a bash command, file edit, or prompt, Claude displays a predefined message or warning.

What events can trigger a Hookify rule?

Hookify supports several event types: 'bash' for terminal commands, 'file' for code edits, 'stop' for when the agent finishes a task, 'prompt' for user input, and 'all' for every event.

Hookify Rule Writing Guide

Name: Hookify Rule Writing Guide
Author: akiojin

byakiojin

•

개발자 도구

Standardizes the creation and configuration of Hookify rules to enforce project-specific guardrails and automated workflows within Claude Code.

This skill provides comprehensive guidance for authoring Hookify rules, which are specialized markdown files that intercept Claude Code's actions to provide warnings, block dangerous commands, or enforce best practices. It helps developers set up automated triggers based on regex patterns in bash commands, file edits, and user prompts, enabling a more secure and disciplined development environment by defining custom project-level guardrails and automated completion checklists.

주요 기능

01Step-by-step guidance for YAML frontmatter syntax and rule structure

02Event-specific configuration for bash commands, file operations, and completion steps

03Integration support for local .claude/ directory organization and gitignore workflows

04Advanced pattern matching using multi-condition regex and field-specific filters

051 GitHub stars

06Best practices for writing effective warning messages and alternative suggestions

사용 사례

01Automatically warning against committing sensitive information like API keys or debug logs

02Enforcing a pre-flight checklist before the AI agent finishes a task

03Preventing accidental execution of destructive shell commands like 'rm -rf'

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add akiojin/llm-router hookify-writing-rules

For use in Claude.ai and ChatGPT

Download Skill