RAG Implementation Skill (wshobson) for Claude Code

Makes Claude design and code retrieval-augmented generation (RAG) apps over your documents, for developers building document Q&A or knowledge chatbots.

Installs: 14.2K; on skills.sh
Stars: 42.1K; on GitHub
Files: 1; besides SKILL.md
Updated: Oct 2026; last repo commit

rag-implementation is a reference skill for building RAG systems: apps that look up your documents first, then answer from them. It lists vector database options, embedding models, retrieval strategies and reranking methods, and gives Claude a Python starter built on LangGraph, Anthropic's Claude model, Voyage embeddings and Pinecone.

It triggers when you build knowledge-grounded AI, document Q&A or semantic search. It bundles one detail file and no scripts. The example code needs Python packages and provider keys such as a Pinecone key, so expect Claude to ask for them.

Good for

Developers building Q&A over proprietary documents
Teams cutting hallucinations with grounded answers and citations
Engineers choosing between Pinecone, Weaviate, Chroma, Qdrant, Milvus or pgvector

Install rag-implementation

Claude Code

Copy this prompt.

Install the rag-implementation skill in Claude Code. Run this command:

/plugin marketplace add wshobson/agents
/plugin install llm-application-dev@claude-code-workflows

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

The plugin is llm-application-dev from the claude-code-workflows marketplace. It installs 8 skills, 3 agents (ai-engineer, prompt-engineer, vector-database-engineer) and 3 slash commands (/llm-application-dev:langchain-agent, :ai-assistant, :prompt-optimize), not only this skill. The plugin folder lists no hooks or MCP config. Pick one route, not both.

From the project README. Source.

What's in the skill

When it loads

Build Retrieval-Augmented Generation (RAG) systems for LLM applications with vector databases and semantic search. Use when implementing knowledge-grounded AI, building document Q&A systems, or integrating LLMs with external knowledge bases.

Sections in SKILL.md

When to Use This Skill
Core Components
1. Vector Databases
2. Embeddings
3. Retrieval Strategies
4. Reranking
Quick Start with LangGraph
Detailed patterns and worked examples

Key facts

Author: wshobson (Seth Hobson)
Repository: wshobson/agents
Installs on skills.sh: 14,194 as of 2026-11-02
GitHub stars: 42,081 for the whole wshobson/agents repo
License: MIT (plugin.json for the llm-application-dev plugin)
Last repo commit: 2026-10-30
What is inside: 2 files including SKILL.md: SKILL.md and references/details.md. It bundles no scripts.
Part of: The llm-application-dev plugin in wshobson/agents, with 8 skills, 3 agents and 3 commands in the repo folder

What it includes

7 use cases, from document Q&A to research tools with citations
Q&A over proprietary documents, chatbots with current facts, semantic search, fewer hallucinations, domain knowledge access, documentation assistants and research tools with source citation.
Try: “Build a document Q&A app over our PDF handbook with citations, using Chroma locally.”
When to use
Pinecone, Weaviate, Milvus, Chroma, Qdrant, pgvector
Six options with a one-line trade-off for each, such as Chroma for local development and pgvector for SQL integration.
Vector databases
Table of 6 models with dimensions and best use
voyage-3-large, voyage-code-3, text-embedding-3-large, text-embedding-3-small, bge-large-en-v1.5 and multilingual-e5-large.
Embedding models
Dense, sparse, hybrid, multi-query and HyDE
Five approaches from plain semantic similarity to generating hypothetical documents.
Retrieval strategies
Cross-encoders, Cohere Rerank, MMR and LLM scoring
Four ways to reorder results for quality or diversity.
Reranking
Retrieve and generate graph with Claude, Voyage and Pinecone
A Python example that wires a retriever and an answer step into a graph. It calls a Claude model, Voyage embeddings and Pinecone.
Write · LangGraph quick start
5 advanced patterns, chunking, vector store setups, evaluation
Hybrid search with rank fusion, multi-query, contextual compression, parent document retrieval and HyDE. Also chunking strategies, store configs for Pinecone, Weaviate, Chroma and pgvector, retrieval tuning, prompts and metrics.
Read · references/details.md

From the official docs

Permissions

Low risk

The skill bundles no scripts, but the code it helps write calls paid services with your keys.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Example code uses paid APIs and keys
Keep keys in environment variables, not in the repo, and check pricing before indexing a large document set. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	No read-only mode is described. The skill bundles SKILL.md and references/details.md and no scripts.
Talks to	The example code calls your LLM provider, Voyage and Pinecone APIs when you run it
Sign-in and keys	Keys are read from environment variables in the example code.
Data retention	Documents you index are sent to the vector store and embedding provider you choose.
Scopes	Claude writes Python code that uses LangGraph, LangChain packages, Voyage embeddings, Pinecone and a Claude model details.md reads PINECONE_API_KEY from the environment The plugin install route also adds 3 agents (ai-engineer, prompt-engineer, vector-database-engineer) and 3 slash commands; the plugin folder lists no hooks or MCP config

Questions

What does the rag-implementation skill do?

It gives Claude a working reference for retrieval-augmented generation: which vector store, embeddings, retrieval and reranking to use, plus a LangGraph starter. Claude uses it when you build document Q&A or semantic search.

How do I install rag-implementation?

Use the install box at the top of this page. It offers the llm-application-dev plugin (8 skills, 3 agents, 3 slash commands) or this skill alone through the skills installer, which the README says installs without the plugin's agents, commands or hooks. Pick one route.

Does it need API keys?

SKILL.md names no key. Its example code uses a Claude model, Voyage embeddings and Pinecone, and references/details.md reads PINECONE_API_KEY. You can swap in other stores and embeddings the skill lists, such as local Chroma and an open-source embedding model.

Which vector database should I pick?

The skill gives one-line guidance: Chroma for lightweight local development, pgvector if you already use PostgreSQL, Pinecone for managed and serverless, Qdrant for filtered search, Weaviate for hybrid search and Milvus for on-premise performance.

Does it work only with Claude?

The starter code uses Anthropic models, but the embedding table also lists OpenAI and open-source models. The repo README lists Claude Code, Codex and Cursor among the supported tools.

Is it safe?

The skill bundles no scripts; it is SKILL.md plus references/details.md. The code it helps Claude write calls paid services such as Pinecone and your LLM provider, so set spend limits on those keys and keep them out of your repo. The plugin route adds 3 agents and 3 slash commands, listed in the install box.

Information

updated Nov 2, 2026

Repository: wshobson/agents
Folder: plugins/llm-application-dev/skills/rag-implementation

wshobson/agents skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 6

wshobson/agents README · github.com
README.md · github.com
marketplace.json · github.com
commands.md · github.com
rag-implementation SKILL.md · github.com
references/details.md · github.com