Rag Retrieval

This is the reference guide you want open when building production RAG systems. It covers the full stack: basic retrieval patterns, contextual retrieval (Anthropic's technique that cuts retrieval failures by 67%), HyDE for vocabulary mismatches, agentic workflows with self-correction, and pgvector hybrid search with proper HNSW indexing. The 30 rules are split across on-demand files, so you're not loading multimodal chunking logic when you just need basic embeddings. What I appreciate is the honest tradeoff guidance: it tells you hybrid search should be 40% BM25 / 60% vector, that HyDE needs a 2-3 second timeout with fallback, and why you'd pick text-embedding-3-small over voyage-3.5. Complexity is marked high because RAG done right involves real decisions about reranking, token budgets, and when to decompose queries.

skills.sh installs: 1.5K
GitHub stars: 301
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add yonatangross/orchestkit --skill rag-retrieval --agent claude-code

Installs into .claude/skills of the current project.

Files

RAG Retrieval

Comprehensive patterns for building production RAG systems. Each category has individual rule files in rules/ loaded on-demand.

House thresholds, fusion ordering, and the latency and quality budgets we assert live in House delta below. Vendor documentation is linked, not restated (see Upstream coverage).