Stage 2 of the Daktus CDSS pipeline: eval and correction harness for LLM-generated clinical protocols. Pydantic schemas, AST-based logic validation, LLM-as-judge audits, multi-model routing via OpenRouter, closed-loop feedback learning, cost telemetry.

Quality layer for Agent Skills (agentskills.io) — lint, hash, sign, version, eval your SKILL.md files. 10 structural checks, Ed25519 signing, semantic diff, 3-stage eval pipeline. Bidirectional SKILL.md ↔ AIF conversion.

Production-grade RAG pipeline with section-based scraping, retrieval tuning, and LLM-as-judge evaluation — built on Pinecone and Claude.