RAG Agent Skill

A reusable, documented RAG pipeline pattern: async ingestion, sentence-aware chunking with overlap, hybrid vector + keyword retrieval, citation-required generation with retry/backoff, and two-tier caching, distilled for reuse on future projects.