How do I persist environment variables across a session?

You can use a SessionStart hook to append export commands to the file path provided by the $CLAUDE_ENV_FILE environment variable.

What is the difference between plugin hooks and user settings?

Plugin hooks use a wrapped JSON format within hooks.json, while user settings in .claude/settings.json use a direct format with events at the top level.

Can I use hooks to block specific commands?

Yes, the PreToolUse hook allows you to define matchers for specific tools and return a 'deny' decision to prevent execution based on your custom security logic.

What are prompt-based hooks in Claude Code?

Prompt-based hooks use LLM-driven decision making to evaluate tool inputs or task completion using natural language reasoning, offering more flexibility than traditional bash scripts.

Do hooks run sequentially or in parallel?

All hooks matching a specific event run in parallel to optimize performance, meaning they do not see each other's output and should be designed for independence.

Hook Development for Claude Code

Name: Hook Development for Claude Code
Author: syahiidkamil

bysyahiidkamil

•

275

•

开发者工具

Implements event-driven automation and validation for Claude Code plugins using prompt-based and command hooks.

This skill provides a comprehensive framework for extending Claude Code's capabilities through event-driven hooks. It allows developers to intercept tool usage, enforce security policies, validate task completion, and inject project-specific context at key points in the AI's workflow. By supporting both LLM-powered prompt hooks for nuanced reasoning and command hooks for deterministic bash scripts, it empowers engineers to build robust, safe, and context-aware AI agent plugins that align with professional standards like KISS and DRY.

主要功能

01Granular event support including PreToolUse, PostToolUse, and SessionStart

02275 GitHub stars

03Context-aware validation using LLM-driven prompt hooks

04Deterministic execution of bash command hooks for performance

05Security enforcement for path traversal and sensitive file access

06Regex-based tool matching for precise intervention control

使用场景

01Automatically loading project context and environment variables at session start

02Blocking dangerous bash commands or sensitive file writes before execution

03Verifying code quality and build status before allowing the agent to finalize a task

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add syahiidkamil/software-engineer-ai-agent-atlas hook-development

For use in Claude.ai and ChatGPT

Download Skill