agentleFS
Sign inStart free
m00kkm00kk/agent-skills-playbook/skills/rag-agent-pipeline/SKILL.md

rag-agent-pipeline

Skill0 starsChanged 6 months agoMITView on GitHub

Builds retrieval-augmented agent pipelines: document ingest, chunking, embedding, vector store, and grounded answers with citations. Use for RAG agents, knowledge bases, pgvector, Chroma, or enterprise doc Q&A.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not chunk tables blindly — use row-aware or CSV path

    line 32
  2. Distinctive:

    Avoid one tool that dumps entire index.

    line 47
  3. Very specific:

    Run eval set on every index or prompt change (see evaluate-agent-quality)

    line 53
  4. Distinctive:

    Do not mix embedding models across index versions without re-embed.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

RAG Agent Pipeline

Pipeline stages

Ingest → Clean → Chunk → Embed → Index → Retrieve → Generate (with citations)

Workflow

- [ ] Choose sources (PDF, MD, HTML, tickets, APIs)
- [ ] Chunk: 512–1024 tokens, overlap 10–15%, preserve headings
- [ ] Embed: one model for index + query (same dimensions)
- [ ] Store: metadata filters (product, date, ACL)
- [ ] Retrieve: hybrid search (vector + keyword) when possible
- [ ] Generate: require citations; refuse if no relevant chunks

Chunking rules

  • Split on headings first, then size
  • Attach metadata: source, page, updated_at, acl_group
  • Do not chunk tables blindly — use row-aware or CSV path

Retrieval

  • Top-k 5–8 for synthesis; rerank to 3 if reranker available
  • Score threshold: below threshold → "I don't have enough context"
  • Filter by user ACL in metadata before LLM sees chunks

Agent tool surface

Expose to agent as MCP or native tools:

  • search_knowledge(query, filters?) → ranked chunks + scores
  • get_document(doc_id) → full section when needed

Avoid one tool that dumps entire index.

Evaluation

  • Golden questions with expected citation sources
  • Metrics: recall@k, faithfulness, answer relevance
  • Run eval set on every index or prompt change (see evaluate-agent-quality)

Stack hints (pick one per project)

StoreWhen
pgvectorAlready on Postgres
Chroma / QdrantGreenfield vector DB
Managed (Pinecone, etc.)Ops offload

Do not mix embedding models across index versions without re-embed.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free