Guide · Agent skill
How to use the RAG Implementation agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The rag-implementation skill is a set of guidance and example code that teaches an AI agent how to build Retrieval-Augmented Generation (RAG) systems. In RAG, a model first searches a collection of documents for relevant passages and then writes an answer based on them. This guide explains what the skill does, how to install it, how to trigger it, and what to check before using it.
RAG Implementation Guidewshobson/agents · MIT Download the skill Quick start
- 1Download and unzip. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 2Or upload in the Claude apps. Go to Settings → Capabilities → Skills and upload the ZIP instead.
- 3Ask a RAG question in plain language. Describe your documents and goal, such as building a document Q&A system.
- 4Choose a vector database and embedding model. Ask the agent to compare options such as Pinecone, Chroma or pgvector and embeddings such as voyage-3-large.
- 5Set up the example environment. Install LangGraph, LangChain, langchain_anthropic, langchain_voyageai and langchain_pinecone, and get the required API keys and Pinecone account.
- 6Load your documents and test. Index your documents into the vector store yourself, then run the graph and check answers against the source passages.
What is an agent skill and what does this one do?
An agent skill is a folder of instructions that an AI agent can load when a task matches the skill's description. This skill, rag-implementation, is written for developers building knowledge-grounded AI, such as document Q&A systems or documentation chatbots.
Once installed, the agent can recommend vector databases such as Pinecone, Chroma, Weaviate, Milvus, Qdrant or pgvector, suggest embedding models, compare retrieval strategies like hybrid search and multi-query retrieval, and suggest reranking methods. It can also write example Python code using LangGraph and LangChain.
Install the rag-implementation skill
Download the ZIP file from this page and unzip it. You can install the skill in one of two ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder.
In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Trigger the skill with natural requests
You do not need a special command. The skill is matched by its description, which covers implementing knowledge-grounded AI, building document Q&A systems, and integrating LLMs with external knowledge bases. Ask for the outcome you want in plain language.
- Help me design a RAG system that answers questions over our internal docs.
- Which vector database should I use for a small document Q&A chatbot, Chroma or pgvector?
- Compare voyage-3-large and text-embedding-3-small for my support articles.
- Write a LangGraph retrieve-then-generate pipeline with citations to source documents.
Use the included LangGraph example
The SKILL.md file includes a quick start that builds a two-node graph. The retrieve node fetches the top documents from a vector store, and the generate node sends them to a Claude model with a prompt that tells it to say so if it cannot answer.
The example expects a Python environment with LangGraph, LangChain, langchain_anthropic, langchain_voyageai and langchain_pinecone. It also needs an Anthropic API key, a Voyage AI API key and a Pinecone account, and it assumes an existing Pinecone index named docs. The skill does not index your files for you; you must load documents into the vector store yourself.
from langgraph.graph import StateGraph, START, END
from langchain_anthropic import ChatAnthropic
from langchain_voyageai import VoyageAIEmbeddings
from langchain_pinecone import PineconeVectorStore
llm = ChatAnthropic(model="claude-sonnet-5")
embeddings = VoyageAIEmbeddings(model="voyage-3-large")
vectorstore = PineconeVectorStore(index_name="docs", embedding=embeddings)
retriever = vectorstore.as_retriever(search_kwargs={"k": 4})Go beyond the basic pipeline
The references/details.md file contains more advanced patterns. Its examples include hybrid search that combines BM25 keyword retrieval with dense embedding retrieval using weights of 0.3 and 0.7, and multi-query retrieval that generates several query variations with an LLM for better recall.
Ask the agent to read the details file when the basic pattern is not enough for your documents.
Safety and privacy notes
This skill runs no scripts and uses no network access on its own. It provides guidance and code for you to run. If you follow the example, your documents are sent to the embedding service and stored in the vector database you configure, so review those providers' data policies before loading sensitive material.
- Do not load private documents into third-party services until you have reviewed their data policies.
- Keep API keys out of source files and shared code.
- Check model names and availability yourself, since the model list is labelled 2026 and may go stale.
What people use it for
- Designing a document Q&A system over internal company documents.
- Building a documentation assistant that cites the pages it used.
- Comparing vector databases such as Chroma and pgvector for a local prototype.
- Choosing an embedding model for a support knowledge base.
- Adding reranking or hybrid search when basic retrieval misses relevant passages.
Frequently asked questions
How do I install the rag-implementation skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills instead.
Is the rag-implementation skill free?
The skill is published under the MIT licence. The services in the example, such as Anthropic, Voyage AI and Pinecone, are separate products that you sign up for and that may have their own costs.
What do I need to run the example code?
You need a Python environment, the LangGraph, LangChain, langchain_anthropic, langchain_voyageai and langchain_pinecone packages, an Anthropic API key, a Voyage AI API key and a Pinecone account. The example also assumes an existing Pinecone index named docs.
Does the skill index my files or connect to my database?
No. The skill provides knowledge and example code only. You need to load your documents into a vector store yourself and configure the connection.
Which vector database should I pick instead of this skill's default?
The skill lists Pinecone, Weaviate, Milvus, Chroma, Qdrant and pgvector. Pinecone is managed and serverless, Chroma suits lightweight local development, and pgvector integrates with PostgreSQL. Ask the agent to compare them for your scale and setup.
Why does the skill say the model list is 2026?
The embedding model table is labelled 2026 and does not include current pricing or availability. Check each provider's documentation before committing to a model.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a developer or student who wants a clear starting architecture for a document Q&A system, or a team comparing vector databases and embedding models. Skip it if you need a finished search tool over your files, need verified current pricing or model availability, or cannot send your documents to third-party embedding or vector services.
Download the skill