mcp-to-skill（MCP TO 技能）

MCP to Skill Converter

Convert MCP servers into Claude Code Skills for easier distribution and usage.

Conversion Workflow

Step 1: Analyze MCP Server

Run the analysis script to extract tool/resource/prompt definitions:

python3 scripts/analyze_mcp.py /path/to/mcp-project --pretty

The script outputs JSON with: tools, resources, prompts, mcp_sdks, dependencies.

For manual analysis, inspect these key files:

• package.json / pyproject.toml — dependencies and entry point
• src/index.ts / main.py — server setup, tool/resource/prompt registration
• Source files — implementation details

Step 2: Map MCP Components to Skill Structure

Map all three MCP primitives:

For the complete mapping table and conversion patterns, see references/MAPPING.md.

Step 3: Choose Conversion Pattern Per Tool

For each MCP tool, pick the right pattern:

• Pattern A (Script): Complex logic, API calls, error handling → scripts/tool_name.py
• Pattern B (Instruction): Simple commands, no deps → inline in SKILL.md
• Pattern C (Hybrid): Tools requiring server runtime → setup instructions + MCP connection

Decision: Can it run standalone? → A or B. Is it trivial? → B. Needs server state? → C.

For detailed pattern descriptions and examples, see references/MAPPING.md. For real-world conversion case studies, see references/EXAMPLES.md.

Step 4: Generate Skill Structure

{skill-name}/
├── SKILL.md                    # Core instructions
├── scripts/
│   └── {tool_name}.{ext}      # Pattern A tool scripts
├── references/
│   └── {topic}.md             # MCP Resources → reference docs
└── config/
    └── secrets.example.md     # Required env vars / API keys

Step 5: Write SKILL.md

---
name: {skill-name}
description: |
  {Original MCP server description}. Use when:
  (1) {Primary use case from tool descriptions}
  (2) {Secondary use case}
---

# {Skill Name}

{Brief description}

## Prerequisites

{Setup: permissions, API keys, runtime requirements}
Required environment variables: (list from MCP server's .env / config)

## Tools

### {Tool Name}
{Description, usage, parameters, example}

Step 6: Handle Secrets & Environment Variables

• Never include actual secrets in skill files
• Create config/secrets.example.md listing all required env vars with descriptions
• Document minimum required scopes/permissions in Prerequisites
• See references/MAPPING.md § "Secrets & Environment Variables" for patterns

Step 7: Verify the Converted Skill

Before packaging, run through this checklist:

• SKILL.md has proper frontmatter (name, description with triggers)
• All MCP Tools are documented with usage instructions
• All MCP Resources are converted to references/ files
• All MCP Prompts are converted to workflow sections
• Scripts are executable (chmod +x) and tested standalone
• config/secrets.example.md exists if env vars are needed
• No content duplication between SKILL.md and references
• No unnecessary files (README, CHANGELOG, LICENSE, etc.)

Test the skill by:

1. Installing it to a test platform directory
2. Invoking each tool/instruction manually
3. Verifying scripts handle missing args and errors gracefully