AgentStack
MCP verified MIT Self-run

Mcp To Uber Skills Converter

mcp-bjornslib-mcp-to-uber-skills-converter · by bjornslib

Claude Code plugin that converts MCP servers to Claude Skills with progressive disclosure. Reduces context usage by 90%+.

No reviews yet
0 installs
0 views
view→install

Install

$ agentstack add mcp-bjornslib-mcp-to-uber-skills-converter

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Mcp To Uber Skills Converter? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

MCP Skills Plugin

Convert MCP servers to Claude Skills with progressive disclosure, reducing context usage by 90%+.

Features

  • MCP to Skill Converter: Transform any MCP server into a context-efficient Claude Skill
  • Session Start Hook: Automatically loads the MCP skills registry on session start
  • Progressive Disclosure: Only load tool definitions when actually needed
  • Multi-Transport Support: Compatible with stdio, http, and sse MCP server types
  • Automatic Registry Management: Maintains an index of all converted skills

Installation

Install the plugin using Claude Code's --plugin-dir option:

# Clone or download the plugin
git clone https://github.com/your-org/mcp-skills-plugin.git

# Run Claude Code with the plugin
claude --plugin-dir /path/to/mcp-skills-plugin

Or add to your Claude Code configuration for persistent use.

Requirements

  • Python 3.8+
  • mcp package: pip install mcp

Usage

Converting MCP Servers to Skills

Once the plugin is installed, Claude can use the mcp-to-skill-converter skill.

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
brave-search              stdio      ✅ Yes        npx

📊 Total: 10 servers, 8 compatible
Convert a Single Server
python ${PLUGIN_DIR}/skills/mcp-to-skill-converter/mcp_to_skill.py --name github

This will:

  1. Create a skill in .claude/skills/mcp-skills/github/
  2. Update the MCP skills registry
  3. Remove the server from .mcp.json to prevent duplicate loading
Convert All Compatible Servers
python ${PLUGIN_DIR}/skills/mcp-to-skill-converter/mcp_to_skill.py --all

Using Converted Skills

After conversion, Claude can invoke the MCP tools through the generated skill:

# List tools in a skill
python .claude/skills/mcp-skills/executor.py --skill github --list

# Get tool schema
python .claude/skills/mcp-skills/executor.py --skill github --describe create_issue

# Call a tool
python .claude/skills/mcp-skills/executor.py --skill github --call '{"tool": "create_issue", "arguments": {"title": "Bug fix", "body": "Details..."}}'

Session Start Hook

The plugin automatically loads the MCP skills registry at the start of each Claude Code session. This provides:

  • Context efficiency: Only ~150 tokens for the registry vs 30-50k tokens for native MCP loading
  • Discovery: Claude knows which skills are available without loading all tool definitions
  • On-demand loading: Full tool details are only loaded when needed

Context Savings

| Scenario | Native MCP | Skills Plugin | Savings | |----------|------------|---------------|---------| | Idle (no tools used) | 30-100k tokens | ~150 tokens | 99%+ | | Using 1 skill | 30-100k tokens | ~5k tokens | 90%+ | | Tool execution | 30-100k tokens | 0 tokens | 100% |

Plugin Structure

mcp-skills-plugin/
├── .claude-plugin/
│   └── plugin.json           # Plugin manifest
├── skills/
│   └── mcp-to-skill-converter/
│       ├── SKILL.md          # Skill documentation
│       ├── mcp_to_skill.py   # Converter script
│       └── templates/
│           ├── index.json
│           └── registry-SKILL.md
├── hooks/
│   ├── hooks.json            # Hook configuration
│   └── load-mcp-skills.sh    # Session start hook
├── README.md
└── LICENSE

Server Compatibility

| Type | Compatible | Notes | |------|------------|-------| | stdio | ✅ Yes | Standard input/output protocol | | http | ✅ Yes | Streamable HTTP protocol | | sse | ✅ Yes | Server-Sent Events protocol |

Security Notes

⚠️ Important Security Considerations:

  1. Command Execution: The converter and executor scripts run MCP servers as subprocesses. Only convert and use MCP servers from trusted sources.
  1. Environment Variables: MCP server configurations may include environment variables (API keys, tokens). These are stored in the generated mcp-config.json files.
  1. Code Execution: Generated skills can execute arbitrary code through MCP tool calls. Review the tools available in each converted skill before use.
  1. File System Access: The converter reads from .mcp.json and writes to the skills directory. Ensure appropriate file permissions.

Troubleshooting

"mcp package not found"

pip install mcp

"Could not find .mcp.json"

Specify the path explicitly:

python mcp_to_skill.py --mcp-json /path/to/.mcp.json --name github

"Server not found"

Run --list to see available servers in your .mcp.json.

Hook not loading

Ensure the hook script is executable:

chmod +x hooks/load-mcp-skills.sh

Contributing

Contributions are welcome! Please read the contributing guidelines before submitting PRs.

License

MIT License - see [LICENSE](LICENSE) for details.


Reduce your Claude Code context usage by 90%+ with MCP Skills.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.