Rag implementation
Skill yigityildiz0/universal-ai-skill-library/skills/common/rag-implementation
531 searchable AI Agent Skills for Claude Code, OpenAI Codex, and OpenCode — EN/TR catalog, platform and risk notes, direct ZIPs, and curated bundles.
npx -y skills add yigityildiz0/universal-ai-skill-library --skill rag-implementationAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
3 things to look at
- 18 days oldThe repository was created 18 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Design, implement, and evaluate retrieval-augmented generation with traceable ingestion, privacy-aware indexing, hybrid retrieval, grounded answers, and.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.2 KB, as published. Nobody here has run it
Retrieval-Augmented Generation
Contract and corpus
Define users/questions, answer/citation requirements, freshness, latency/cost, permissions, deletion, languages, and failure policy. Inventory source ownership, formats, versioning, sensitive data, access controls, and update cadence. Do not index content the application is not authorized to reveal.
Pipeline
- Parse deterministically and preserve source ID, version/hash, page/section/line, title, ACL, and timestamps.
- Chunk on semantic/structural boundaries with limited overlap; keep tables/code and parent context intentionally.
- Build a lexical baseline. Add embeddings only when evaluation shows benefit; keep backend/model configurable and record exact versions.
- Filter by authorization before retrieval and again before answer assembly.
- Retrieve bounded candidates, combine lexical/semantic ranks when useful, rerank a small set, diversify duplicates, and pass only relevant evidence.
- Instruct the answer layer to use provided evidence, cite source IDs/ranges, distinguish inference, and abstain when support is insufficient.
- Incrementally update by content hash and remove stale/deleted chunks and caches.
External embedding/vector/reranking services require authorization and a privacy/retention review. Do not install infrastructure or send private corpora outside the approved boundary automatically.
Evaluation
Create labeled queries with relevant sources and expected abstentions. Measure retrieval recall@k/MRR or nDCG, citation correctness, answer support, unsupported-claim rate, ACL leakage, freshness, latency, cost, and failure recovery. Compare no-retrieval, lexical, semantic, and hybrid baselines. Test injection inside documents, conflicting sources, stale updates, deleted documents, and permission changes.
Completion report
Document corpus/provenance, parser/chunker, index and versioning, ACL/deletion, retrieval/reranking, prompt/output contract, evaluation results, observability, deployment/rollback, and unresolved coverage.