rag-architect

An agent skill by Jeffallan, from Jeffallan/claude-skills. Tags: database, evaluation, ml, research, search.

What it does

Designs and implements production-grade RAG systems by chunking documents, generating embeddings, configuring vector stores, building hybrid search pipelines, applying reranking, and evaluating retrieval quality. Use when building RAG systems, vector databases, or knowledge-grounded AI applications requiring semantic search, document retrieval, context augmentation, similarity search, or embedding-based indexing.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add Jeffallan/claude-skills --skill rag-architect

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/Jeffallan/claude-skills
cp -r claude-skills/skills/rag-architect ~/.claude/skills/rag-architect

Safety box score

Grade B, low risk: 23 of 100 risk points, rated 2026-10-20. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill is mostly architecture guidance and code examples for building RAG systems, not a runnable tool. The main risks are that its examples encourage network calls to third-party APIs/services, package installs, and handling API keys or local data stores if copied into a real project.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	1 (low)	The files contain runnable Python/SQL examples and commented install commands such as `# pip install ragas` and `# pip install trulens-eval`, so an agent could be led to execute code even though the skill itself ships no scripts.
Network access	2 (moderate)	Examples call external services and download content, including `OpenAI()`, `cohere.Client(...)`, `weaviate.connect_to_weaviate_cloud(...)`, `QdrantClient(url=...)`, and `nltk.download('punkt')`, so using the guidance can send prompts/data off machine.
Filesystem reach	1 (low)	The guidance writes to local paths like `./chroma_data`, `.embedding_cache`, and `./fine-tuned-model`, which is limited to the working directory but still creates persistent files.
Secrets & credentials	1 (low)	It explicitly references API-key handling such as `os.environ["COHERE_API_KEY"]` and placeholder keys for OpenAI/Pinecone/Cohere, so it touches credentials even though it advises using env vars or a secrets manager.
Privilege & persistence	0 (none)	There is no sudo, service registration, shell startup editing, or other privilege-changing behavior in the skill files.
Instruction hijack surface	0 (none)	I found no attempts to override the agent's rules, hide information, fetch remote instructions, or use prompt-injection tactics.
Irreversible actions	0 (none)	The skill describes design and evaluation steps only; it does not instruct the agent to deploy, publish, message, pay, or otherwise take irreversible external actions.
Transparency	1 (low)	Some examples rely on opaque runtime dependencies, especially `AutoModel.from_pretrained(..., trust_remote_code=True)` and model/package downloads from external registries, which makes part of the executed code/data path unreviewable at install time.

Before installing

Read the examples that use external services (`OpenAI`, `Cohere`, Weaviate, Qdrant Cloud) and confirm you are okay sending prompts/documents off machine.
If you plan to use the late-chunking example, review `trust_remote_code=True` in `references/chunking-strategies.md` before running it.
Install only in a virtualenv after checking the commented package installs in `references/rag-evaluation.md` (`ragas`, `trulens-eval`).
Confirm the local output paths the examples use (`.embedding_cache`, `./chroma_data`, `./fine-tuned-model`) are acceptable in your workspace.

Rated 2026-10-20 by an automated review of 6 files from the skill's repository.

Source

Repository: Jeffallan/claude-skills (all skills from this repository)
Path: skills/rag-architect/SKILL.md
Branch: main
Updated: 2026-09-19

storing-and-querying-vectors — Store and query vector embeddings using Amazon S3 Vectors, a cost-effective long-term vector storage service with its own API namespace (s3vectors).
customer-research — When the user wants to conduct, analyze, or synthesize customer research.
prospecting — When the user wants to find, qualify, and build a list of prospects to reach out to — across B2B SaaS, general B2B, or local small businesses.
fine-tuning-expert — Use when fine-tuning LLMs, training custom models, or adapting foundation models for specific tasks.
ml-pipeline — Designs and implements production-grade ML pipeline infrastructure: configures experiment tracking with MLflow or Weights & Biases.
the-fool — Use when challenging ideas, plans, decisions, or proposals using structured critical reasoning.