A retrieval-gated skill architecture for LLM agents that scales to hundreds of tools by exposing only the top-K relevant capabilities per request.

Architecture-first virtual task branching for agentic work

Keyword-first skill routing and topology layer for file-based agent systems.

AI Agent Semantic Routing Framework - Multi-layer semantic retrieval for intelligent skill dispatch

A Codex skill that routes users to the best local skills, keeps the chosen skill active across the conversation, and asks before setup or downloads.

Fix Gemma 4, Llama 4, Qwen 3, Mistral tool-calling failures. Zero-latency regex skill router for Ollama & OpenClaw — 400,000x faster than LLM inference, 100% accurate. No GPU needed.

Local MCP server providing flat search routing for 200+ professional skills. Fuzzy search across all skills, load full skill body on demand.

Meridian MCP — calibrated AI skill routing via orbital mechanics