MCP to Skill Converter

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

Conversion Workflow

1. Analyze MCP Server

Read and understand the MCP server structure:

bash

# Key files to analyze
- package.json / pyproject.toml  # Dependencies and entry point
- src/index.ts / main.py         # Entry point and tool registration
- src/**/*.ts / **/*.py          # Tool implementations

Extract this information:

•Server name and description
•Available tools (name, description, parameters, implementation)
•Dependencies (runtime requirements)
•Execution method (node, python, etc.)

2. Map MCP Tools to Skill Structure

MCP Concept	Skill Equivalent
Tool name	Script or instruction section
Tool description	Used in SKILL.md description
Tool parameters	Script arguments or instruction parameters
Tool implementation	`scripts/` executable or inline instructions

3. Generate Skill Structure

code

{skill-name}/
├── SKILL.md                    # Core instructions
├── scripts/
│   ├── run_server.sh          # Server startup script (optional)
│   └── {tool_name}.{ext}      # Individual tool scripts
└── references/
    └── tools.md               # Tool reference documentation

4. Write SKILL.md

Template structure:

markdown

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

# {Skill Name}

{Brief description of what this skill does}

## Prerequisites

{Any setup requirements - permissions, API keys, etc.}

## Available Tools

{List each tool with usage instructions}

### {Tool Name}

{Description}

**Usage:**
{How to invoke - either via script or direct instruction}

**Parameters:**
- `{param}`: {description}

**Example:**
{Concrete usage example}

Conversion Patterns

Pattern A: Script-based (for complex tools)

When MCP tool has complex logic, create executable script:

python

# scripts/tool_name.py
#!/usr/bin/env python3
import argparse
# ... implementation

Reference in SKILL.md:

markdown

Run `scripts/tool_name.py --param value`

Pattern B: Instruction-based (for simple tools)

When MCP tool is simple, use inline instructions:

markdown

### Send Notification

To send a system notification:
1. Use AppleScript: `display notification "message" with title "title"`

Pattern C: Hybrid (server-dependent tools)

When tools require the MCP server runtime:

markdown

## Setup

Start the MCP server:
\`\`\`bash
cd {project-path}
npm start  # or: node dist/index.js
\`\`\`

Then use MCP tools via the running server.

AppleScript MCP Example

For applescript-mcp specifically:

•No server needed - Tools are standalone AppleScripts
•Use instruction-based pattern - Each tool becomes a section
•Group by category - System, Calendar, Finder, etc.
•Include permission notes - macOS security requirements

Output Checklist

Before packaging, verify:

• SKILL.md has proper frontmatter (name, description)
• Description includes all use cases and triggers
• All MCP tools are documented
• Scripts are executable and tested
• References are complete but not redundant
• No unnecessary files (README, CHANGELOG, etc.)

Package the Skill

bash

python3 ~/.claude/skills/skill-creator/scripts/package_skill.py /path/to/skill