MCP-to-Skill Converter
Installation
The skill invokes
.claude/tools/integrations/mcp-converter/batch_converter.py
. Requirements:
- Python 3.10+: python.org or
winget install Python.Python.3.12
- pip: Usually included with Python; verify with .
- Dependencies: From the repo root, install deps for the integration (e.g. PyYAML if required):
Run from project root; the script uses
.claude/tools/integrations/mcp-converter/
Cheat Sheet & Best Practices
MCP design: Single responsibility per server; bounded toolsets; contracts first (strict I/O schemas); stateless by default; additive changes; security (identity, auth, audit). Prefer stdio for local, Streamable HTTP for remote; use a gateway for multi-tenant/centralized policy.
Conversion: Introspect server; estimate token usage of tool schemas; generate skill with progressive disclosure. Test converted skills before relying on them. Use catalog + batch_converter for rules-driven conversion.
Hacks: Focus on high-token or high-value servers first. Keep generated SKILL.md and wrappers in version control. Use
to mark
or auto-convert thresholds.
Certifications & Training
MCP: MCP Best Practices,
modelcontextprotocol.info.
Skill data: Single responsibility, bounded tools, contracts first, stateless; stdio vs HTTP; gateway pattern; introspect → generate skill.
Hooks & Workflows
Suggested hooks: Post–MCP config change: optional batch_converter run to refresh skills. Use with evolution-orchestrator (add mcp-converter to secondary) when creating skills from MCP servers.
Workflows: Use with
evolution-orchestrator. Flow: list servers → convert server or batch → test converted skill. See
creators/skill-creator-workflow.yaml
; mcp-converter feeds skill-creator input.
🚀 Usage
1. List Available MCP Servers
See which servers are configured in your
:
bash
python .claude/tools/mcp-converter/mcp_analyzer.py --list
2. Convert a Server
Convert a specific MCP server to a Skill:
bash
python .claude/tools/mcp-converter/mcp_analyzer.py --server <server_name>
3. Batch Conversion (Catalog)
Convert multiple servers based on rules:
bash
python .claude/tools/mcp-converter/batch_converter.py
ℹ️ How it Works
- Introspect: Connects to the running MCP server.
- Analyze: Estimates token usage of tool schemas.
- Generate: Creates a wrapper that creates dynamic tool calls only when needed.
🔧 Dependencies
Memory Protocol (MANDATORY)
Before starting:
Read
.claude/context/memory/learnings.md
After completing:
- New pattern ->
.claude/context/memory/learnings.md
- Issue found ->
.claude/context/memory/issues.md
- Decision made ->
.claude/context/memory/decisions.md
ASSUME INTERRUPTION: If it's not in memory, it didn't happen.