claude-hooks-developer

创建、配置和管理Claude Code钩子以实现工作流自动化、验证和安全。指导钩子实现、配置模式和最佳实践。

📖 中文导读:创建、配置和管理Claude Code钩子以实现工作流自动化、验证和安全。指导钩子实现、配置模式和最佳实践。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: claude-hooks-developer description: Create, configure, and manage Claude Code hooks for workflow automation, validation, and security. Guides hook implementation, configuration patterns, and best practices.

Claude Code Hooks Developer

Compact skill for creating and managing Claude Code hooks that intercept tool calls, validate operations, and enhance workflows.

When to Use

Activate when the user:

  • Needs to create a new hook (PreToolUse, PostToolUse, UserPromptSubmit, Stop, etc.)
  • Wants to configure hooks in .claude/settings.json or .claude/settings.local.json
  • Asks about hook lifecycle events or available hook types
  • Needs to validate/block operations, format files, or send notifications
  • Mentions security controls, file protection, or command validation
  • Wants examples of hook patterns or common use cases

Core Capabilities

  • Hook Types: Guide selection between PreToolUse, PostToolUse, UserPromptSubmit, Stop, SubagentStop, SessionStart, Notification, PermissionRequest
  • Hook Configuration: Generate correct JSON structure for settings.json with matchers, commands, and paths
  • Script Templates: Provide Python/Bash templates for common hook patterns
  • Security Patterns: Implement file protection, dangerous command blocking, and access control
  • Environment Variables: Utilize CLAUDE_PROJECT_DIR, CLAUDE_TOOL_INPUT_*, and other hook context
  • Decision Control: Properly use exit codes (0=approve, 2=block) and JSON decision objects
  • Common Use Cases: Format files, validate commands, send notifications, log operations, enforce policies

Quick Workflow

  1. Identify the hook event type based on the desired trigger point
  2. Create the hook script in .claude/hooks/ (Python or Bash)
  3. Make script executable (chmod +x)
  4. Add hook configuration to .claude/settings.json or .claude/settings.local.json
  5. Test the hook with relevant tool operations
  6. Verify proper exit codes and error messages

Hook Event Types

  • PreToolUse: Runs before tool execution (can block/modify)
  • PostToolUse: Runs after tool execution (validation/cleanup)
  • UserPromptSubmit: Validates user prompts before processing
  • Stop: Controls whether Claude continues working
  • SubagentStop: Controls sub-agent continuation
  • SessionStart: Runs at session initialization
  • Notification: Responds to system notifications
  • PermissionRequest: Auto-approves/denies permission dialogs

Exit Codes

  • 0: Approve/Success - operation continues
  • 2: Block - operation blocked with error message
  • Other: Error - operation fails

Reference Documentation

  • Detailed Workflows & Patterns: See reference.md in this directory
  • Real-World Usage Examples: See examples.md in this directory
  • Hook Configuration Guide: .claude/hooks/doc-standards-reminder.sh (example)