{"id":77097,"plugin_id":"plugins_6ab42163e9048191ab6e8c14258a48c3","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-11-03T00:12:26.325Z","digest":"7da1010298ac7f95905b786975e1d2d8a46c955713e6c74d5806a5bf9706b453","against":null,"payload":{"description":"Design and optimize search, metadata filters, hybrid retrieval, reranking, query rewriting and context selection.","included_files":[],"name":"design-retrieval","skill_md_contents":"---\nname: design-retrieval\ndescription: Design and optimize search, metadata filters, hybrid retrieval, reranking, query rewriting and context selection.\n---\n\n\n# Design retrieval and ranking\n**Trigger:** Poor search relevance, keyword mismatch, top-k instability, reranker regressions or retrieval stack trade-offs.\n\n1. Define target relevance from labeled query/document examples; include hard negatives, exact identifiers, paraphrases and no-answer questions.\n2. Inspect lexical, vector and hybrid candidates separately; record which required evidence is absent at candidate generation vs reranking.\n3. Check query rewrite/routing, metadata filters, ACL constraints, score interpretation and reproducible retrieval parameters.\n4. Compare candidate recall@k, rank-sensitive NDCG/MRR and context utilization on the same held-out set. Keep retrieval evaluation separate from generation.\n5. Only add rerankers, expansion, hybrid fusion or extra retrievers where an ablation supports net value; account for latency and cost.\n6. Ensure no reranker or query rewrite can bypass source-level authorization; preserve source IDs through context packing.\n7. Return proposed search flow, tuned controls, failure-mode tests, evidence metrics and rollback criterion.\n\nRead `../../references/retrieval-ranking.md` and `../../references/evaluation-metrics.md`.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}