What does this skill do?

Karpathy LLM Wiki packages a repeatable workflow for building a personal or team knowledge base that an agent can maintain over time. Instead of re-searching raw documents for every question, the agent compiles sources into durable markdown wiki pages.

The skill focuses on three core operations: ingesting sources into raw/, querying synthesized wiki pages with citations, and linting the wiki for broken links, missing index entries, stale cross-references, and related issues.

  • Ingests web pages, papers, blog posts, PDFs, markdown files, text files, and pasted text

  • Compiles source material into maintained markdown wiki pages

  • Answers questions with citations back to wiki content

  • Checks wiki structure, links, index integrity, and knowledge-base health

How it works

1

Ingest a source

The agent receives a URL, file, or pasted text and stores the source material under the wiki's raw knowledge area.

2

Compile durable wiki pages

The skill helps the agent summarize, cross-reference, and update markdown wiki pages that can improve as new sources arrive.

3

Answer with citations

When asked a question, the agent searches the compiled wiki and grounds its answer in the existing markdown pages.

4

Lint and maintain the wiki

The workflow checks index integrity, broken links, stale references, and other issues so the knowledge base stays useful.

Best use cases for this skill

  • Personal research wikis: Build a durable knowledge base for papers, articles, notes, and technical references.

  • Engineering knowledge management: Help agents maintain project or architecture knowledge without relying only on raw document retrieval.

  • Cited Q&A workflows: Let agents answer recurring questions with links back to synthesized wiki pages.

Frequently asked questions

What is the Karpathy LLM Wiki skill?

Karpathy LLM Wiki is an open-source skill that helps agents ingest sources, compile durable markdown wiki pages, answer questions with citations, and lint the wiki for quality.

How is an LLM wiki different from RAG?

An LLM wiki stores curated markdown pages that are synthesized during ingest and maintenance. RAG usually retrieves raw chunks at query time.

What sources can it ingest?

It can work with web pages, papers, blog posts, PDFs, markdown files, text files, and pasted text, then compile that material into wiki pages.

Start in Buda

Start with this skill.

Add it to an agent and use it in a real workflow when you are ready.

Explore marketplace

Install workflow

  1. 1Choose an agent
  2. 2Attach the skill
  3. 3Use it in a repeatable workflow