Claude Code 技能与钩子：自定义工作流扩展

来源：code.claude.com/docs/en/skills · hooks-guide | 整理时间：2026-04-04

概述

Claude Code 的 Skills 和 Hooks 是两大扩展机制：

• Skills（技能）：将重复性的复杂工作流封装成可复用的"技能"，通过 /command 一键调用
• Hooks（钩子）：在 Claude Code 执行工具操作的前后自动触发自定义脚本

两者结合，可以把 Claude Code 从一个通用助手定制成专属于你团队的工作流引擎。

Skills 技能系统

什么是 Skill？

Skill 是一个 .md 文件，包含 YAML 元数据和使用说明。Claude 会自动发现并加载这些技能。

<!-- .claude/skills/fix-issue/SKILL.md -->
---
name: fix-issue
description: Fix a GitHub issue end-to-end
---

1. Use `gh issue view` to get the issue details
2. Search the codebase for relevant files
3. Create a plan for the fix
4. Implement the fix
5. Write and run tests to verify
6. Create a descriptive commit and PR referencing the issue

使用方式：/fix-issue 1234

Skill 的目录结构

.claude/skills/
├── fix-issue/
│   └── SKILL.md          ← 修复 GitHub issue 的技能
├── code-review/
│   └── SKILL.md          ← 代码审查技能
├── deploy-check/
│   └── SKILL.md          ← 部署前检查
└── generate-test/
    └── SKILL.md          ← 自动生成测试

自动发现机制

• Claude 会扫描 .claude/skills/ 目录下所有 SKILL.md 文件
• 支持 monorepo：子目录中的技能仅在该目录上下文中激活
• 技能优先级高于普通的 .claude/commands/ 自定义命令

实用技能示例

示例 1：代码审查技能

<!-- .claude/skills/code-review/SKILL.md -->
---
name: review
description: Comprehensive code review with security focus
---

1. Read the current git diff (staged and unstaged)
2. Analyze for:
   - Logic errors and edge cases
   - Security vulnerabilities (injection, XSS, auth)
   - Performance concerns
   - Code style consistency with CLAUDE.md rules
3. Rate severity: 🔴 Critical / 🟡 Warning / 🔵 Suggestion
4. Provide fix suggestions for each finding

示例 2：数据库迁移技能

<!-- .claude/skills/migrate/SKILL.md -->
---
name: migrate
description: Generate and run a database migration
---

1. Understand the schema change from user description
2. Generate both UP and DOWN migration files
3. Review for data loss risks
4. Run the migration in development
5. Verify the schema is correct

社区技能资源

Hooks 钩子系统

三种钩子类型

四个生命周期事件

配置方式

在 .claude/settings.json 中配置：

{
  "hooks": {
    "PostToolUse": [
      {
        "type": "shell",
        "command": "npx eslint --fix $CLAUDE_FILE_PATH",
        "matcher": "Edit|Write"
      }
    ],
    "PreToolUse": [
      {
        "type": "shell",
        "command": "echo 'About to modify: $CLAUDE_FILE_PATH'",
        "matcher": "Edit|Write"
      }
    ]
  }
}

实用钩子示例

自动 lint（PostToolUse）

{
  "hooks": {
    "PostToolUse": [
      {
        "type": "shell",
        "command": "npx prettier --write $CLAUDE_FILE_PATH 2>/dev/null",
        "matcher": "Edit|Write"
      }
    ]
  }
}

安全审查（PreToolUse）

{
  "hooks": {
    "PreToolUse": [
      {
        "type": "prompt",
        "prompt": "Review this tool call for security risks. Block if it: deletes production data, exposes secrets, or modifies critical infrastructure.",
        "matcher": "Bash"
      }
    ]
  }
}

通知到 Slack（Notification）

{
  "hooks": {
    "Notification": [
      {
        "type": "shell",
        "command": "curl -X POST -H 'Content-type: application/json' --data '{\"text\":\"Claude Code: $CLAUDE_NOTIFICATION\"}' $SLACK_WEBHOOK_URL"
      }
    ]
  }
}

自动测试（PostToolUse）

{
  "hooks": {
    "PostToolUse": [
      {
        "type": "agent",
        "prompt": "If the modified file is a source file (not test), find and run the related test file. If tests fail, report the failures.",
        "matcher": "Edit|Write"
      }
    ]
  }
}

让 Claude 帮你写钩子

你不需要手写钩子配置，直接告诉 Claude：

Create a hook that automatically runs prettier on files after each edit

Set up a PreToolUse hook that blocks any Bash command containing 'rm -rf' or 'DROP TABLE'

Claude 会自动生成正确的 settings.json 配置。

Skills + Hooks 组合模式

模式 1：技能激活钩子

{
  "hooks": {
    "PostToolUse": [
      {
        "type": "agent",
        "prompt": "If the user just ran /deploy-check skill and all checks passed, remind them to update CHANGELOG.md",
        "matcher": "Edit|Write"
      }
    ]
  }
}

模式 2：TDD 工作流

{
  "hooks": {
    "PostToolUse": [
      {
        "type": "shell",
        "command": "cd $PROJECT_ROOT && npm test -- --related $CLAUDE_FILE_PATH --watch=false 2>&1 | tail -20",
        "matcher": "Edit|Write"
      }
    ]
  }
}

写测试 → Claude 实现 → 自动运行测试 → 反馈结果，形成闭环。

模式 3：自动安全审查

{
  "hooks": {
    "PreToolUse": [
      {
        "type": "prompt",
        "prompt": "Check for: SQL injection, XSS, command injection, hardcoded secrets, and insecure defaults. If found, block the tool call.",
        "matcher": "Edit|Write"
      }
    ]
  }
}

最佳实践

相关链接

• 最佳实践
• 常见工作流
• MCP 集成
• 多 Agent 协作
• 官方技能文档：https://code.claude.com/docs/en/skills
• 官方钩子文档：https://code.claude.com/docs/en/hooks-guide
• 社区技能合集：https://github.com/hesreallyhim/awesome-claude-code