Local Skills Mcp

This is a filesystem-based skill loader that scans local directories for tools and prompts without jamming everything into context upfront. Instead of registering hundreds of skills at startup, it uses lazy loading to keep the initial context small and discovers capabilities on demand. You'd reach for this when you're building a library of reusable AI skills stored as files and need Claude to access them without burning through context tokens listing every single one. The context-efficient discovery means you can maintain a large collection of tools locally while only surfacing what's relevant to the current task. It's designed for developers who want to organize MCP skills as code in their filesystem rather than hardcoding them into a single server implementation.