MCP to Skill Converter

MCP to Skill Converter

Convert any MCP server into a Claude Skill with progressive disclosure pattern, reducing context usage by 90%+.

When to Use This Skill

• Converting MCP servers to skills for context efficiency
• Listing available MCP servers in a project
• Batch converting multiple MCP servers
• Creating skills from .mcp.json configuration

Quick Commands

All commands run from project root where the plugin is installed.

List Available Servers

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --list

Shows all MCP servers with compatibility status (stdio, http, and sse types are compatible).

Convert Single Server (auto-outputs to mcp-skills/)

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --name <server-name>

Example:

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --name github
# Outputs to: .claude/​skills/​mcp-skills/​github/
# Automatically removes github from .mcp.json
# Updates mcp-skills/​index.json

Convert All Compatible Servers

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --all
# Converts all compatible servers (stdio, http, sse) to mcp-skills/

Custom Output Directory

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --name github --output-dir /​custom/​path

Specify Custom .mcp.json

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --mcp-json /​path/​to/​.mcp.json --name github --output-dir ./​skills

CLI Options

Server Compatibility

Context Savings

Workflow Example

Step 1: List available servers

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --list

Output:

📄 Servers in: /​path/​to/​.mcp.json

Name                      Type       Compatible   Command
--------------------------------------------------------------------------------
github                    stdio      ✅ Yes        npx
context7                  stdio      ✅ Yes        npx
livekit-docs              http       ❌ No         https://...

📊 Total: 16 servers, 14 compatible

Step 2: Convert desired server

python ${PLUGIN_DIR}/​skills/​mcp-to-skill-converter/​mcp_to_skill.py --name github

The converter automatically:

• Creates the skill in .claude/​skills/​mcp-skills/​github/
• Initializes the mcp-skills registry (SKILL.md + index.json) if needed
• Updates index.json with the new skill
• Removes the server from .mcp.json to avoid duplicate loading

Step 3: Add semantic trigger keywords (MANDATORY)

The converter generates a placeholder description. You MUST update it with meaningful trigger keywords based on what the MCP server actually does.

1. Read the generated SKILL.md to understand what tools are available
2. Identify semantic keywords - What would users naturally ask for?
   • Think about user intent, not tool names
   • "ToolUI", "chat components" NOT "assistantUIDocs"
   • "create PR", "issues", "repository" NOT "create_pull_request"
3. Update the SKILL.md description with trigger keywords:

# Before (auto-generated):
description: Dynamic access to assistant-ui MCP server (2 tools)

# After (agent-enhanced):
description: Use for assistant-ui documentation, ToolUI, generative UI, chat components, Thread, Composer, Message primitives, runtime integrations. Get docs and code examples for building AI chat interfaces.

1. Update the registry SKILL.md (mcp-skills/​SKILL.md):
   • Add the skill to the "Available Skills" table
   • Include the trigger keywords column

Example registry entry:

| Skill | Tools | Trigger Keywords |
|-------|-------|------------------|
| assistant-ui | 2 | ToolUI, generative UI, chat components, assistant-ui docs |
| github | 26 | PR, issues, repository, commits, code search |

Step 4: Use the generated skill Claude will auto-discover the new skill in the mcp-skills registry and can invoke MCP tools with minimal context overhead.

Generated Skill Structure

output-dir/​server-name/
├── SKILL.md           # Instructions and tool documentation
├── executor.py        # Async MCP client wrapper
├── mcp-config.json    # Server configuration
└── package.json       # Dependency info

Using Generated Skills

Use the central executor from project root:

# List tools in a skill
python .claude/​skills/​mcp-skills/​executor.py --skill <skill-name> --list

# Describe specific tool
python .claude/​skills/​mcp-skills/​executor.py --skill <skill-name> --describe tool_name

# Call a tool
python .claude/​skills/​mcp-skills/​executor.py --skill <skill-name> --call '{"tool": "tool_name", "arguments": {...}}'

Requirements

pip install mcp

Python 3.8+ required.

Error Handling

Files in This Skill

• SKILL.md - This documentation
• mcp_to_skill.py - The converter script
• templates/ - Template files for generated skills

---

This skill enables converting MCP servers to Claude Skills for 90%+ context savings.