Skrills is a modern Rust-based MCP server and CLI tool designed to empower users with local control over their AI skills, specifically `SKILL.md` files. It streamlines the process of discovering, filtering, and injecting these skills into AI prompts, such as those used by Codex CLI or Claude, while rigorously optimizing for token overhead. By employing intelligent filtering based on prompt content, embeddings, and byte limits, Skrills ensures efficient prompt usage and prevents the inclusion of unnecessarily large skill sets. It also provides manifest-first payloads and detailed diagnostics, offering robust observability and governance for AI client integrations.
Key Features
01Intelligent prompt and embedding filtering for efficient skill inclusion
02Manifest-first payloads for detailed skill summaries and content
033 GitHub stars
04Comprehensive diagnostics and observability for skill injection
05Manual and heuristic auto-pinning of frequently used skills
06Runtime configuration overrides and Claude skills mirroring
