Skip to content
Get packs

Guide · Agent skill

How to use the LlamaIndex RAG Builder agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The llamaindex skill gives your AI agent a practical reference for LlamaIndex, an open-source Python framework that connects language models to your own documents. It is written for Python developers and data teams who want retrieval-augmented generation (RAG) over private files. After this guide you will know what the skill changes, how to install it, how to ask for LlamaIndex code in plain language, and what to check before running the examples.

LlamaIndex RAG BuilderOrchestra-Research/AI-Research-SKILLs · MIT

Quick start

  1. 1Download the ZIP. Get the llamaindex skill ZIP from this page and unzip it.
  2. 2Place the folder. Move it to ~/.claude/skills/ for personal use, or to a project's .claude/skills/ folder.
  3. 3Set up Python. Install the llama-index package and any integrations you need.
  4. 4Add your API key. Provide an OpenAI key or another LLM provider key for the examples.
  5. 5Make a natural request. Describe your documents and the answer you want, naming LlamaIndex if helpful.
  6. 6Review and run the code. Check which data leaves your machine, then run the generated script yourself.

What the llamaindex skill makes the agent do

Agent skills are folders of instructions and reference files that an AI agent loads when a task matches them. This skill's main file, SKILL.md, covers LlamaIndex's core concepts, and supporting reference files cover agents, data connectors and query engines.

Once installed, the agent can write code that loads files or web pages, splits them into searchable pieces, builds an index, and answers questions over that material. It also covers building simple agents that call tools and chatbots that remember earlier turns.

  • Load data with connectors such as SimpleDirectoryReader, web page readers and a GitHub repository reader
  • Build indices, including VectorStoreIndex for semantic search
  • Create query engines and retrievers, with options such as response modes and metadata filters
  • Write agents that wrap a query engine as a tool alongside functions like a calculator
  • Set up chat engines for multi-turn conversations

Install the llamaindex skill

Download the ZIP from the skill's page in this directory and unzip it. The folder contains SKILL.md and a references folder. Place it in one of two locations: your personal skills folder, or a project's own skills folder so it applies only to that project.

You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.

  • Personal: unzip into ~/.claude/skills/
  • Single project: unzip into the project's .claude/skills/ folder

Know what you need before you start

The skill writes code; you still need a Python environment and the LlamaIndex packages. The source installation options are the starter package llama-index, or llama-index-core plus specific integrations such as llama-index-llms-openai and llama-index-embeddings-openai.

The examples call hosted model services. You need an OpenAI API key or another LLM provider key; the examples use OpenAI and Anthropic. Optional connector packages such as llama-index-readers-web or llama-index-readers-github may need their own credentials.

  • A Python environment
  • The llama-index pip package and integrations
  • An OpenAI API key or other LLM provider key
  • Optional connector packages and their credentials
pip install llama-index

Trigger the skill with natural requests

You do not need special syntax. Describe the outcome you want in plain language that mentions documents, search, chat or LlamaIndex, and the agent will use the skill when the request matches its description. Naming LlamaIndex directly makes this more reliable.

Good requests say what the data is, where it lives and what you want back.

  • Use LlamaIndex to answer questions over the PDFs in my ./data folder.
  • Build a RAG pipeline over our Markdown docs and show me how to persist the index.
  • Write a LlamaIndex chatbot that remembers earlier turns of the conversation.
  • Create a LlamaIndex agent that can search my docs and also do arithmetic.

Example: a 5-line RAG script

The core pattern is load, index, query. The skill's basic example loads a folder, builds a vector index and asks one question. Replace the folder name and question with your own. The example needs an LLM key configured in your environment.

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader documents = SimpleDirectoryReader("data").load_data() index = VectorStoreIndex.from_documents(documents) query_engine = index.as_query_engine() response = query_engine.query("What did the author do growing up?") print(response)

Safety notes: scripts, network and private data

This skill does not run scripts. It uses the network: the code it writes calls hosted model APIs, and web or GitHub connectors fetch content from the internet. Anything you send to a hosted API may leave your machine.

Review the code before running it, especially if your documents are sensitive. Check which provider receives the text, and confirm your organisation allows that before loading confidential files.

  • Sensitive documents sent to third-party APIs may leave your machine; review first
  • Connector packages may need credentials; store them in environment variables rather than in code
  • Confirm which provider receives your data

Tips for better results

Choose the response mode to match the job. The source describes compact as the default and suggests tree summarisation for hierarchical summaries. Use metadata filters when your documents have categories, and persist the index so you do not rebuild it on every run.

If another tool fits better, the skill itself points to alternatives: LangChain for more general-purpose agents, Haystack for production search pipelines, txtai for lightweight semantic search, and Chroma when you only need vector storage.

  • Use persist to save an index and load_index_from_storage to reopen it
  • Use similarity_top_k to control how many chunks are retrieved
  • Ask the agent to add verbose output while debugging

What people use it for

  • Document Q&A over a folder of PDFs or Markdown notes
  • An internal knowledge-base chatbot that remembers earlier questions
  • Retrieval over a GitHub repository or documentation website
  • A tool-using agent that searches docs and performs calculations

Frequently asked questions

What is the LlamaIndex skill?

It is an agent skill containing a reference guide to LlamaIndex, an open-source Python framework for connecting language models to your own documents. It helps the agent write correct LlamaIndex code for loading, indexing and querying data.

What do I need to use the llamaindex skill?

You need a Python environment, the llama-index package and any integrations you use, plus an OpenAI API key or another LLM provider key for the examples. Some optional connectors need their own credentials.

Is LlamaIndex free?

The source describes LlamaIndex as open source under the MIT licence, and the skill itself is MIT licensed. Hosted model APIs you call from the code are billed separately by their providers.

Does the skill run code on my computer?

No, the skill does not run scripts. It uses the network, because the code it helps you write calls hosted model APIs and some connectors fetch content online.

Is LlamaIndex better than LangChain?

The skill describes LlamaIndex as best for data-centric applications and document retrieval, while LangChain is more general-purpose and better for agents. Pick based on whether your project centres on documents or on agent workflows.

Can I use this skill with sensitive documents?

Only after review. The examples send content to hosted services such as OpenAI, so check your provider's data handling and your organisation's policy before loading confidential files.

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

This skill suits Python developers and data teams building document Q&A, knowledge-base chatbots or RAG prototypes who want the agent to write LlamaIndex code accurately. It is not a finished application, and if your project needs a different stack or your documents cannot leave your environment, consider the alternatives it names or a local setup.

Download the skill