AGENT SKILL
vector-search-quality
Improve vector retrieval quality with chunk strategy, metadata filters, thresholds, and reranking.
- License 许可
- MIT
- 最近更新
- 2026-05-28
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- vector-search-quality
- author
- Siming
- license
- MIT
- version
- 1.0.0
- metadata
- { "siming": { "tags": [ "rag", "vector", "retrieval" ] } }
- description
- Improve vector retrieval quality with chunk strategy, metadata filters, thresholds, and reranking.
03 SKILL.md 正文
Vector Search Quality
Use this skill when search results are irrelevant or unstable.
Workflow
- Inspect chunk size, overlap, metadata, and query rewriting.
- Tune score thresholds before adding complex rerankers.
- Keep tenant and permission filters mandatory.
- Return retrieval diagnostics for debugging.
- Add examples that compare expected and actual chunks.
Checks
- Retrieval cannot leak data across tenants.
04 同仓库更多 Skills
playwright-visual-check
Use Playwright to verify UI layout, screenshots, responsive behavior, and console/runtime errors.
★ 0
api-smoke-tests
Create API smoke tests for auth, CRUD, pagination, empty states, and error paths.
★ 0
secret-handling
Handle secrets, API keys, tokens, and credentials safely in code, logs, configs, and UI.
★ 0
dependency-risk-scan
Evaluate dependency risk, vulnerability scan results, package updates, and safe upgrade strategy.
★ 0
rag-ingestion-pipeline
Build RAG document ingestion flows with upload, parsing, chunking, embedding, queueing, and status tracking.
★ 0
kb-admin-workflows
Design knowledge base admin workflows for list pages, status filters, import progress, and document operations.
★ 0