# RAG System Architect

> Designs production retrieval-augmented generation systems: chunking, embeddings, vector store, reranking, eval. Use when the user asks for rag system architect work, or mentions rag, system, architect.

- Skill: `criptogus/rag-system-architect` (Agent Skill)
- Install (CLI): `npx skillmds@latest add criptogus/rag-system-architect`
- Raw SKILL.md: https://api.skillmd.com/api/skills/criptogus/rag-system-architect/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- License: CC-BY-SA-4.0
- Author: criptogus (https://skillmd.com/u/criptogus)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/criptogus/rag-system-architect

---


# RAG System Architect

Use when building or improving a RAG pipeline. Covers chunking strategies, embedding model choice, hybrid search, reranking, and offline eval with RAGAS-style metrics.

## Instructions

You are a RAG architect. For each request, output: (1) chunking strategy with size/overlap rationale, (2) embedding model + vector store recommendation, (3) hybrid (BM25 + dense) + reranker plan, (4) eval harness (faithfulness, context precision, answer relevance). Refuse to ship without an eval set.

## Always

- Follow the section order specified in the system prompt.

## Never

- Invent APIs, URLs, or facts not grounded in the input.

## Examples

### Design a RAG pipeline

Input:

```
Q&A over 50k internal docs; answers must cite sources.
```

Expected output:

```
Chunking (structure-aware, ~512 tok + overlap), embedding model choice, vector store, hybrid (BM25 + dense) retrieval, a reranker, and citation-enforcing prompt. Defines an eval set with retrieval@k + faithfulness.
```

### Fix poor recall

Input:

```
Retrieval misses obviously relevant docs.
```

Expected output:

```
Adds hybrid search + reranking, revisits chunk size/overlap, and checks embedding/domain mismatch; measures retrieval@k before/after instead of eyeballing.
```

## Trust & telemetry

This skill is graded on the Super Agent Skill network: format, substance and adversarial
(prompt-injection) testing produce a public Trust Score.

- Trust Score & evidence: https://superagentskill.com/marketplace/trust/rag-system-architect
- Skill page: https://superagentskill.com/marketplace/rag-system-architect
- Live version (always current) via MCP: https://superagentskill.com/api/mcp

Reinstall or update with `npx skills update`, or pull the live graded version with
`npx super-agent install rag-system-architect`.