Claude Code Plugins生态完整指南：从安装到自定义开发

✅ 术语表（3分钟） - 快速了解关键概念
✅ 第2章:5分钟快速开始（15分钟）
✅ 第3.1-3.2节:浏览和安装（10分钟）
✅ 第5.1节:快速验证（2分钟）

✅ 第1章:Plugins简介（20分钟）
✅ 第2章:快速开始（15分钟）
✅ 第4章:创建自定义Plugin（90分钟）- 重点章节
✅ 第5.3节:发布到Marketplace（30分钟）
✅ 第6章:故障排查（25分钟）

🔧 第6章:故障排查 - 按错误类型查找
🔧 第7章:FAQ - 最常见的20个问题

手机              |  Claude Code
------------------|------------------
操作系统(iOS/Android) | Claude Code核心
App Store        | Plugin Marketplace
安装的APP        | 已安装的Plugins
APP更新          | Plugin自动更新

### 1. 确认Claude Code已安装
claude --version
### 预期输出:claude-code/2.1.12 或更高版本（2025年1月后）

### 2. 确认在项目目录中
cd /path/to/your/project
ls .claude/  # 应该看到settings.json等文件

{
  "marketplaces": {
    "enabled": true,
    "autoUpdate": true,
    "sources": [
      "https://code.claude.com/plugins/index.json"
    ]
  }
}

### 列出所有可用Plugins
claude plugins list --available

### 搜索PDF相关Plugins
claude plugins search pdf

🔍 搜索结果: 3个Plugins匹配 "pdf"

📦 document-skills:pdf (v1.2.0)
   📝 Comprehensive PDF manipulation toolkit
   ⭐ 4.8/5.0 (1,234 安装)
   🏷️  Tags: pdf, document, extract

📦 pdf-analyzer (v0.9.0)
   📝 Advanced PDF content analysis
   ⭐ 4.2/5.0 (456 安装)

📦 invoice-pdf-parser (v1.0.3)
   📝 Extract data from invoice PDFs
   ⭐ 3.9/5.0 (234 安装)

### 安装document-skills:pdf
claude plugins install document-skills:pdf

### 或使用完整URL安装
claude plugins install https://github.com/anthropics/document-skills

📥 正在下载 document-skills:pdf v1.2.0...
✅ 下载完成

🔍 验证Plugin schema...
✅ Schema有效（符合Anthropic 2025规范）

📦 安装依赖:
   - @anthropic/pdf-lib v2.1.0 ✅
   - pdfjs-dist v3.11.174 ✅

🔧 配置Plugin hooks...
   - pre-commit: pdf-validation ✅

✅ document-skills:pdf 安装成功!

💡 使用方法:
   claude "Extract text from example.pdf"
   或在对话中提到PDF文件

### 查看已安装Plugins
claude plugins list

### 输出:
### 📦 已安装Plugins (1):
###    ✅ document-skills:pdf v1.2.0

### 启动Claude Code
claude

### 在对话中输入:
> Extract the first page from my-document.pdf and save as page1.pdf

### 卸载Plugin
claude plugins uninstall document-skills:pdf

### 输出:
### 🗑️  正在卸载 document-skills:pdf...
### ✅ 已删除文件: .claude/plugins/document-skills/
### ✅ 已清理依赖: @anthropic/pdf-lib
### ✅ 卸载完成

### 列出所有分类
claude marketplace categories

### 输出:
### 📂 可用分类:
###    - document-processing (45个Plugins)
###    - code-quality (32个Plugins)
###    - data-analysis (28个Plugins)
###    - api-integration (67个Plugins)
###    - project-management (21个Plugins)
###    - visualization (18个Plugins)
###    - testing (15个Plugins)
###    - other (17个Plugins)

### 查看某个分类
claude marketplace category document-processing

┌─────────────────────────────────────────┐
│ Claude Code Plugin Marketplace          │
│                                         │
│ 🔍 [搜索框: "搜索Plugins..."]          │
│                                         │
│ 📂 分类:                                │
│  ├─ 📄 文档处理 (45)                    │
│  ├─ 🔧 代码质量 (32)                    │
│  ├─ 📊 数据分析 (28)                    │
│  └─ 🔌 API集成 (67)                     │
│                                         │
│ 🌟 热门Plugins:                         │
│  ┌──────────────────────────┐          │
│  │ document-skills:pdf      │          │
│  │ ⭐⭐⭐⭐⭐ 4.8 (1,234)     │          │
│  │ [安装] [详情]            │          │
│  └──────────────────────────┘          │
└─────────────────────────────────────────┘

### 按名称安装
claude plugins install document-skills:pdf

### 指定版本
claude plugins install document-skills:pdf@1.2.0

### 安装最新测试版
claude plugins install document-skills:pdf@beta

### 完整仓库URL
claude plugins install https://github.com/anthropics/document-skills

### 特定分支
claude plugins install https://github.com/user/my-plugin#dev-branch

### 特定commit
claude plugins install https://github.com/user/my-plugin#abc123f

### 从本地路径安装
claude plugins install /path/to/my-plugin

### 使用相对路径
cd my-plugin/
claude plugins install .

### 跳过依赖安装（快速测试）
claude plugins install my-plugin --no-deps

### 强制重新安装
claude plugins install my-plugin --force

### 全局安装（所有项目可用）
claude plugins install my-plugin --global

### 静默安装（无进度输出）
claude plugins install my-plugin --silent

### 简洁列表
claude plugins list

### 详细信息
claude plugins list --verbose

### 输出示例:
📦 已安装Plugins (3):

1. document-skills:pdf v1.2.0
   📁 路径: .claude/plugins/document-skills/
   📅 安装时间: 2025-12-15 10:23:45
   🔄 更新状态: 最新版本
   📊 使用次数: 47次

2. task-master-ai v2.1.0
   📁 路径: .claude/plugins/task-master-ai/
   📅 安装时间: 2025-12-10 14:12:30
   ⚠️  更新状态: v2.2.0可用
   📊 使用次数: 123次

3. code-review-expert v1.0.5
   📁 路径: .claude/plugins/code-review/
   📅 安装时间: 2025-12-08 09:45:12
   🔄 更新状态: 最新版本
   📊 使用次数: 89次

### 更新单个Plugin
claude plugins update task-master-ai

### 更新所有Plugins
claude plugins update --all

### 检查可用更新（不实际更新）
claude plugins outdated

### 输出:
### 📦 可更新Plugins (1):
###    task-master-ai: v2.1.0 -> v2.2.0
###    Changelog: https://github.com/task-master/releases/v2.2.0

### 临时禁用（不卸载）
claude plugins disable document-skills:pdf

### 重新启用
claude plugins enable document-skills:pdf

### 查看禁用状态
claude plugins list --show-disabled

### 显示Plugin元数据
claude plugins info document-skills:pdf

### 输出:
📦 document-skills:pdf v1.2.0

📝 描述:
   Comprehensive PDF manipulation toolkit for extracting text,
   tables, creating new PDFs, merging/splitting documents.

👤 作者: Anthropic Agent Skills Team
🔗 仓库: https://github.com/anthropics/document-skills
📜 License: MIT
⭐ 评分: 4.8/5.0 (1,234 安装)

🏷️  Tags: pdf, document, extract, merge, split

🔧 提供的工具:
   - extract_pdf_text: 提取文本
   - extract_pdf_tables: 提取表格
   - create_pdf: 创建新PDF
   - merge_pdfs: 合并PDF
   - split_pdf: 拆分PDF

📋 依赖:
   - @anthropic/pdf-lib ^2.1.0
   - pdfjs-dist ^3.11.174

🔗 相关Plugins:
   - pdf-analyzer: 高级PDF分析
   - ocr-plugin: PDF OCR识别

{
  "plugins": {
    "enabled": true,
    "autoUpdate": true,
    "updateCheckInterval": 86400,  // 24小时检查一次
    "allowList": [],  // 空数组=允许所有
    "denyList": ["untrusted-plugin"],
    "marketplace": {
      "sources": [
        "https://code.claude.com/plugins/index.json",
        "https://raw.githubusercontent.com/jeremylongshore/claude-code-plugins-plus/main/index.json"
      ]
    }
  }
}

{
  "document-skills:pdf": {
    "enabled": true,
    "config": {
      "maxFileSize": "50MB",
      "cacheEnabled": true,
      "defaultDPI": 300,
      "ocrLanguage": "eng+chi_sim"
    }
  }
}

my-plugin/
├── plugin.json          # 必需: Plugin元数据
├── README.md            # 推荐: 使用文档
├── skills/              # 可选: Agent Skills
│   └── my-skill/
│       ├── SKILL.md
│       └── prompts/
├── commands/            # 可选: Slash Commands
│   └── my-command.md
├── hooks/               # 可选: Hooks
│   └── pre-commit.py
└── mcp/                 # 可选: MCP配置
    └── mcp-config.json

{
  "schema": "https://schemas.anthropic.com/claude-code/plugin/v1",
  "name": "my-awesome-plugin",
  "displayName": "My Awesome Plugin",
  "version": "1.0.0",
  "description": "A plugin that does awesome things",
  "author": {
    "name": "Your Name",
    "email": "you@example.com",
    "url": "https://github.com/yourname"
  },
  "license": "MIT",
  "repository": {
    "type": "git",
    "url": "https://github.com/yourname/my-plugin"
  },
  "keywords": ["utility", "automation"],
  "engines": {
    "claudeCode": ">=2.0.0"
  },
  "capabilities": {
    "skills": true,
    "commands": true,
    "hooks": true,
    "mcp": false
  },
  "dependencies": {
    "@anthropic/skill-utils": "^1.0.0"
  },
  "config": {
    "schema": {
      "type": "object",
      "properties": {
        "apiKey": {
          "type": "string",
          "description": "Your API key"
        }
      }
    },
    "default": {
      "apiKey": ""
    }
  }
}

### 创建Plugin目录
mkdir -p my-first-plugin/commands
cd my-first-plugin

{
  "schema": "https://schemas.anthropic.com/claude-code/plugin/v1",
  "name": "hello-world-plugin",
  "displayName": "Hello World Plugin",
  "version": "1.0.0",
  "description": "A simple hello world plugin",
  "author": {
    "name": "Claude Student"
  },
  "license": "MIT",
  "capabilities": {
    "commands": true
  }
}

Say hello to the user in a friendly way.

Steps:
1. Get the current user's name from the system
2. Get the current time
3. Respond with a personalized greeting

Example output:
"Hello, Alice! 👋 It's 10:23 AM. How can I help you today?"

### Hello World Plugin

A simple plugin that demonstrates Plugin creation.

### Features

- `/hello` command for friendly greetings

### Installation

\`\`\`bash
claude plugins install /path/to/hello-world-plugin
\`\`\`

### Usage

\`\`\`bash
claude
> /hello
\`\`\`

### 从当前目录安装
claude plugins install .

### 测试命令
claude
> /hello

### 预期输出:
Hello, Alice! 👋 It's 10:23 AM. How can I help you today?

### 初始化Git仓库
git init
git add .
git commit -m "Initial commit: Hello World Plugin"

### 推送到GitHub
git remote add origin https://github.com/yourname/hello-world-plugin.git
git push -u origin main

claude plugins install https://github.com/yourname/hello-world-plugin

pdf-watermark-plugin/
├── plugin.json
├── README.md
├── LICENSE
├── skills/
│   └── pdf-watermark/
│       ├── SKILL.md
│       ├── prompts/
│       │   ├── watermark-instructions.md
│       │   └── batch-watermark.md
│       └── tools/
│           └── watermark.py
├── commands/
│   └── add-watermark.md
└── examples/
    └── sample-usage.md

{
  "schema": "https://schemas.anthropic.com/claude-code/plugin/v1",
  "name": "pdf-watermark",
  "displayName": "PDF Watermark Plugin",
  "version": "1.0.0",
  "description": "Add text/image watermarks to PDF files",
  "author": {
    "name": "PDF Tools Team",
    "email": "team@pdftools.com"
  },
  "license": "MIT",
  "repository": {
    "type": "git",
    "url": "https://github.com/pdftools/watermark-plugin"
  },
  "keywords": ["pdf", "watermark", "document", "branding"],
  "engines": {
    "claudeCode": ">=2.0.0"
  },
  "capabilities": {
    "skills": true,
    "commands": true
  },
  "dependencies": {
    "@anthropic/skill-utils": "^1.0.0",
    "pypdf": "^3.17.0",
    "pillow": "^10.1.0"
  },
  "config": {
    "schema": {
      "type": "object",
      "properties": {
        "defaultText": {
          "type": "string",
          "description": "Default watermark text",
          "default": "CONFIDENTIAL"
        },
        "defaultOpacity": {
          "type": "number",
          "description": "Watermark opacity (0-1)",
          "default": 0.3,
          "minimum": 0,
          "maximum": 1
        },
        "defaultPosition": {
          "type": "string",
          "description": "Watermark position",
          "enum": ["center", "top-right", "bottom-right", "diagonal"],
          "default": "diagonal"
        }
      }
    }
  }
}

### PDF Watermark Skill

Add text or image watermarks to PDF files with customizable positioning, opacity, and styling.

### Capabilities

1. **Text Watermarks**: Add customizable text overlays
2. **Image Watermarks**: Add logo/image watermarks
3. **Batch Processing**: Watermark multiple PDFs at once
4. **Position Control**: Center, corners, diagonal placement
5. **Opacity Control**: Transparent to opaque watermarks

### Usage Examples

#### Add text watermark
\`\`\`
Add "DRAFT" watermark to contract.pdf, diagonal placement, 30% opacity
\`\`\`

#### Add image watermark
\`\`\`
Add company-logo.png to all PDFs in /documents, top-right corner
\`\`\`

#### Batch watermark
\`\`\`
Add "CONFIDENTIAL" to all PDFs in current folder, save to /watermarked
\`\`\`

### Configuration

See plugin config in `.claude/plugins/pdf-watermark/config.json`

#!/usr/bin/env python3
"""
PDF Watermark Tool
Add text/image watermarks to PDF files
"""
import sys
import json
from pathlib import Path
from typing import Literal

try:
    from pypdf import PdfReader, PdfWriter
    from PIL import Image, ImageDraw, ImageFont
except ImportError:
    print("Error: Required packages not installed")
    print("Run: pip install pypdf pillow")
    sys.exit(1)

def add_text_watermark(
    input_pdf: str,
    output_pdf: str,
    text: str = "WATERMARK",
    position: Literal["center", "top-right", "bottom-right", "diagonal"] = "diagonal",
    opacity: float = 0.3,
    font_size: int = 50
) -> bool:
    """
    Add text watermark to PDF

    Args:
        input_pdf: Input PDF file path
        output_pdf: Output PDF file path
        text: Watermark text
        position: Placement position
        opacity: Transparency (0-1)
        font_size: Font size in points

    Returns:
        True if successful
    """
    try:
        reader = PdfReader(input_pdf)
        writer = PdfWriter()

        for page in reader.pages:
            # Get page dimensions
            page_width = float(page.mediabox.width)
            page_height = float(page.mediabox.height)

            # Create watermark image
            img = Image.new('RGBA', (int(page_width), int(page_height)), (255, 255, 255, 0))
            draw = ImageDraw.Draw(img)

            # Calculate position
            font = ImageFont.load_default()
            text_bbox = draw.textbbox((0, 0), text, font=font)
            text_width = text_bbox[2] - text_bbox[0]
            text_height = text_bbox[3] - text_bbox[1]

            if position == "center":
                x = (page_width - text_width) / 2
                y = (page_height - text_height) / 2
            elif position == "top-right":
                x = page_width - text_width - 20
                y = 20
            elif position == "bottom-right":
                x = page_width - text_width - 20
                y = page_height - text_height - 20
            elif position == "diagonal":
                x = page_width / 4
                y = page_height / 2
                img = img.rotate(45, expand=True)

            # Draw text with opacity
            color = (128, 128, 128, int(255 * opacity))
            draw.text((x, y), text, fill=color, font=font)

            # Convert to PDF and overlay
            # ... (完整实现省略,实际需要更多代码)

            writer.add_page(page)

        with open(output_pdf, 'wb') as f:
            writer.write(f)

        return True

    except Exception as e:
        print(f"Error: {e}", file=sys.stderr)
        return False

if __name__ == "__main__":
    # CLI入口
    if len(sys.argv) < 3:
        print("Usage: watermark.py <input.pdf> <output.pdf> [text] [position] [opacity]")
        sys.exit(1)

    input_pdf = sys.argv[1]
    output_pdf = sys.argv[2]
    text = sys.argv[3] if len(sys.argv) > 3 else "WATERMARK"
    position = sys.argv[4] if len(sys.argv) > 4 else "diagonal"
    opacity = float(sys.argv[5]) if len(sys.argv) > 5 else 0.3

    success = add_text_watermark(input_pdf, output_pdf, text, position, opacity)

    if success:
        print(f"✅ Watermark added: {output_pdf}")
    else:
        print("❌ Failed to add watermark")
        sys.exit(1)

Add a watermark to a PDF file.

Parameters:
- $1: Input PDF file path
- $2: Watermark text (optional, default: "WATERMARK")
- $3: Position (optional: center|top-right|bottom-right|diagonal, default: diagonal)
- $4: Opacity (optional: 0.0-1.0, default: 0.3)

Steps:
1. Validate input PDF exists
2. Load plugin configuration for default settings
3. Call skills/pdf-watermark/tools/watermark.py with parameters
4. Verify output PDF was created
5. Show success message with output path

Example:
/add-watermark contract.pdf "DRAFT" diagonal 0.5

### 安装Plugin
cd pdf-watermark-plugin
claude plugins install .

### 测试命令
claude
> /add-watermark test.pdf "CONFIDENTIAL" center 0.4

### 预期输出:
✅ Watermark added: test_watermarked.pdf

{
  "version": "1.2.3"
  //        │ │ │
  //        │ │ └─ Patch: Bug修复
  //        │ └─── Minor: 向后兼容的新功能
  //        └───── Major: 破坏性变更
}

### Plugin Name

一句话描述

### Features
- 功能1
- 功能2

### Installation
\`\`\`bash
安装命令
\`\`\`

### Usage
\`\`\`bash
使用示例
\`\`\`

### Configuration
配置说明

### Troubleshooting
常见问题

### License

### ❌ 错误示例:直接抛出异常
def process_pdf(file):
    reader = PdfReader(file)  # 可能失败
    return reader

### ✅ 正确示例:友好错误提示
def process_pdf(file):
    try:
        if not Path(file).exists():
            print(f"❌ 文件不存在: {file}")
            return None

        reader = PdfReader(file)
        return reader

    except Exception as e:
        print(f"❌ 处理PDF失败: {e}")
        print("💡 提示: 确认文件未损坏")
        return None

### 在Plugin初始化时验证配置
def validate_config(config):
    required_fields = ["apiKey", "maxFileSize"]
    for field in required_fields:
        if field not in config:
            raise ValueError(f"Missing required config: {field}")

    if config["maxFileSize"] > 100 * 1024 * 1024:  # 100MB
        print("⚠️  Warning: maxFileSize > 100MB may cause performance issues")

{
  "dependencies": {
    "@anthropic/skill-utils": "^1.0.0",  // 使用^允许小版本更新
    "pypdf": "~3.17.0"  // 使用~只允许patch更新
  },
  "peerDependencies": {
    "claudeCode": ">=2.0.0"  // 明确最低版本要求
  }
}

✅ plugin.json符合Anthropic 2025 Schema
✅ README.md包含安装和使用说明
✅ LICENSE文件（推荐MIT/Apache 2.0）
✅ 所有依赖在package.json/requirements.txt中声明
✅ 版本号遵循语义化版本规范
✅ 测试通过（至少在本地测试）

⭐ CHANGELOG.md记录版本变更
⭐ examples/目录包含使用示例
⭐ 截图/GIF展示Plugin效果
⭐ GitHub Topics标签（如claude-code-plugin, pdf, automation）

### 使用官方工具验证
npx @anthropic/plugin-validator validate plugin.json

### 输出示例:
✅ Schema valid (v1)
✅ Required fields present
⚠️  Warning: No examples/ directory found (recommended)
✅ Dependencies declared correctly

### 本地初始化
cd my-plugin
git init
git add .
git commit -m "Initial release v1.0.0"

### 创建远程仓库（在GitHub网站上创建）
### 然后:
git remote add origin https://github.com/yourname/my-plugin.git
git branch -M main
git push -u origin main

claude-code-plugin
claude-code
automation
(其他相关标签)

### 1. 检查名称拼写
claude plugins search my-plugin

### 2. 刷新Marketplace索引
claude plugins refresh

### 3. 使用完整URL安装
claude plugins install https://github.com/author/my-plugin

❌ Error: Dependency conflict
   my-plugin@1.0.0 requires skill-utils@^2.0.0
   other-plugin@1.5.0 requires skill-utils@^1.0.0

### 方法1:更新冲突的Plugin
claude plugins update other-plugin

### 方法2:安装兼容版本
claude plugins install my-plugin@0.9.0

### 方法3:使用--force强制安装（可能导致问题）
claude plugins install my-plugin --force

❌ Error: Failed to install dependency '@anthropic/pdf-lib'
   npm ERR! 404 Not Found

### 1. 检查npm配置
npm config get registry
### 应该是: https://registry.npmjs.org/

### 2. 清理npm缓存
npm cache clean --force

### 3. 手动安装依赖
cd .claude/plugins/my-plugin/
npm install

### 4. 检查Node.js版本
node --version  # 应该>=18.0.0

> /my-command
❌ Unknown command: /my-command

### 1. 确认Plugin已安装
claude plugins list

### 2. 检查Plugin是否启用
claude plugins info my-plugin

### 3. 重启Claude Code
claude --restart

### 4. 查看日志
tail -f ~/.claude/logs/plugins.log

### 给脚本添加执行权限
chmod +x .claude/plugins/my-plugin/hooks/pre-commit.sh

### 检查文件所有者
ls -la .claude/plugins/my-plugin/hooks/

### 1. 检查Python版本
python --version  # 应该>=3.8

### 2. 安装依赖
pip install -r .claude/plugins/my-plugin/requirements.txt

### 3. 检查虚拟环境
which python  # 确认使用正确的Python

### 4. 使用插件内置的虚拟环境
cd .claude/plugins/my-plugin
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

// .claude/plugins/my-plugin/config.json
{
  "maxFileSize": "100MB"
}

### 1. 检查配置文件路径
ls -la .claude/plugins/my-plugin/config.json

### 2. 验证JSON格式
jq . .claude/plugins/my-plugin/config.json

### 3. 重启Claude Code使配置生效
claude --restart

### 4. 查看Plugin读取的配置
claude plugins config my-plugin

### 方法1:在启动前设置
export MY_PLUGIN_API_KEY="your-key"
claude

### 方法2:在.claude/settings.json中配置
{
  "env": {
    "MY_PLUGIN_API_KEY": "${env:MY_PLUGIN_API_KEY}"
  }
}

### 方法3:Plugin配置文件中设置
{
  "my-plugin": {
    "apiKey": "your-key"  // 不推荐:明文存储
  }
}

### 设置日志级别
export CLAUDE_LOG_LEVEL=debug

### 启动Claude Code
claude

### 实时查看
tail -f ~/.claude/logs/plugins.log

### 查看特定Plugin日志
grep "my-plugin" ~/.claude/logs/plugins.log

### 查看错误日志
grep "ERROR" ~/.claude/logs/plugins.log

### 直接运行Python脚本
cd .claude/plugins/my-plugin
python tools/my-tool.py --help

### 调试模式运行
python -m pdb tools/my-tool.py

### 查看Plugin统计
claude plugins stats my-plugin

### 输出:
📊 Plugin Statistics: my-plugin
   - Total invocations: 47
   - Total tokens used: 12,345
   - Average tokens per call: 262
   - Last used: 2025-12-15 10:23

claude plugins info my-plugin | grep "requiresNetwork"

### 列出所有Plugin
claude plugins list

### 批量卸载
claude plugins list --format=names | xargs -n1 claude plugins uninstall

### 或手动删除
rm -rf .claude/plugins/*

### 方法1:使用全局+项目级隔离
claude plugins install my-plugin@1.0.0 --global
cd project-a
claude plugins install my-plugin@2.0.0  # 项目级

### 方法2:使用别名
claude plugins install my-plugin@1.0.0 --alias my-plugin-v1
claude plugins install my-plugin@2.0.0 --alias my-plugin-v2

{
  "permissions": {
    "network": {
      "enabled": true,
      "allowedDomains": [
        "api.example.com",
        "*.github.com"
      ]
    }
  }
}

⚠️  my-plugin requests network access to:
   - api.example.com
   - *.github.com

Allow? [y/N]

### ✅ 允许:读取设置
settings = read_claude_settings()

### ❌ 禁止:直接修改核心设置
### 需要用户确认

### 建议用户修改,而不是自动修改
def suggest_config_change():
    print("💡 建议在.claude/settings.json中添加:")
    print(json.dumps({"myPlugin": {"enabled": true}}, indent=2))

my-plugin/
├── plugin.json
├── skills/
├── tests/
│   ├── test_basic.py
│   └── test_integration.py
└── README.md

import pytest
from skills.my_skill.tools import my_function

def test_my_function():
    result = my_function("input")
    assert result == "expected output"

def test_error_handling():
    with pytest.raises(ValueError):
        my_function(None)

cd .claude/plugins/my-plugin
pytest tests/

{
  "pluginDependencies": {
    "document-skills:pdf": "^1.0.0",
    "data-viz-pro": ">=2.0.0"
  }
}

### 启用Plugin调试模式
export CLAUDE_PLUGIN_DEBUG=true

### 启动Claude Code
claude

### 在对话中使用Plugin时,会输出详细日志:
> /my-command test

🔍 [Plugin Debug] my-plugin
   - Command: /my-command
   - Args: ["test"]
   - Context: {...}
   - Execution time: 234ms

{
  "mcp": {
    "servers": {
      "my-tool-server": {
        "command": "node",
        "args": ["./mcp-server.js"]
      }
    }
  }
}

import { Server } from "@modelcontextprotocol/sdk/server/index.js";

const server = new Server({
  name: "my-tool-server",
  version: "1.0.0"
}, {
  capabilities: {
    tools: {}
  }
});

server.setRequestHandler("tools/list", async () => ({
  tools: [{
    name: "my_custom_tool",
    description: "Does something useful",
    inputSchema: {
      type: "object",
      properties: {
        input: { type: "string" }
      }
    }
  }]
}));

### 启用性能分析
claude plugins profile my-plugin

### 运行Plugin
> /my-command

### 查看报告
claude plugins profile report my-plugin

### 输出:
📊 Performance Report: my-plugin
   Function           Calls    Total Time    Avg Time
   ─────────────────────────────────────────────────
   process_pdf()      12       2.34s         195ms
   extract_text()     48       1.12s         23ms
   save_output()      12       0.45s         38ms