メニュー
Route RAG retrieval quality work across hybrid search, reranking, query transformation, HyDE, Self-RAG, RAPTOR, CRAG, and Graph RAG.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use this skill when building, debugging, or improving Retrieval-Augmented Generation systems, including chunking, vector database selection, hybrid search, reranking, multimodal RAG, code documentation RAG, retrieval latency, and production RAG architecture.
Chunk nested documents into parent-child levels so retrieval can move from broad sections to fine-grained passages.
Use semantic boundaries and embedding similarity to chunk text for higher-relevance retrieval.
Route RAG chunking decisions across semantic, hierarchical, sliding-window, contextual-header, and framework-selection strategies.
Use overlapping windows to preserve context across chunk boundaries while controlling retrieval size.
Reduce retrieval latency with caching, batching, and index-level optimization.
| name | retrieval-strategies |
| title | Retrieval Strategies |
| description | Route RAG retrieval quality work across hybrid search, reranking, query transformation, HyDE, Self-RAG, RAPTOR, CRAG, and Graph RAG. |
| category | retrieval-strategies |
| tags | ["retrieval","ranking","hybrid-search","rag"] |
| allowed-tools | ["Read","Grep","Glob"] |
Use this parent skill when the main RAG problem is search quality, ranking, recall, context selection, or evidence traceability. Route to the child skill that best matches the retrieval failure.
Embedding search alone often misses exact terms, ranks weak evidence too high, or retrieves incomplete context. RAG systems need retrieval strategies that adapt to query type and evidence requirements.
Identify whether the issue is recall, ranking, query mismatch, missing context, or poor citations.
Choose hybrid search for exact-term misses, reranking for noisy top-k results, query transformation for vague questions, and corrective patterns for unreliable evidence.
Measure recall at k, MRR or nDCG, citation accuracy, answer faithfulness, and latency impact.
