karpathy-llm-wiki
Agent Skills-compatible LLM wiki for Claude Code, Cursor, and Codex. Build a Karpathy-style knowledge base from raw sources, citations, and linting.
What is karpathy-llm-wiki?
Agent Skills-compatible LLM wiki for Claude Code, Cursor, and Codex. Build a Karpathy-style knowledge base from raw sources, citations, and linting.
Overview of karpathy-llm-wiki
What it is
karpathy-llm-wiki is an Agent Skills-compatible skill for building a Karpathy-style LLM wiki with Claude Code, Cursor, and Codex. It packages the LLM wiki idea into an installable skill that lets a coding agent ingest sources into a raw/ directory, compile durable knowledge pages into a wiki/ directory, answer questions with citations, and lint the wiki for consistency. An LLM wiki is a knowledge system where the model maintains structured wiki pages instead of re-searching raw documents on every query. New sources are compiled into durable markdown pages, cross-references are updated over time, and answers cite the wiki pages that already contain synthesized knowledge.
Key features
- Ingest: Collects a source into
raw/, triages it, then creates or updates wiki articles — or logs it when nothing is new. - Query: Searches the wiki and answers with citations linking to markdown pages.
- Lint: Checks index integrity, links, and wiki health, with auto-fixes and reported issues.
- Source support: Accepts web pages, papers, blog posts, PDFs, markdown files, text files, and pasted text.
- Tool compatibility: Works with Claude Code, Cursor, Codex CLI, OpenCode, and other tools supporting the Agent Skills standard.
Use cases
Building a compounding knowledge base from raw sources, summaries, and durable cross-links. Maintaining a wiki that improves over time instead of re-deriving relationships on every query. Suitable for projects where a human focuses on choosing sources and asking questions while the LLM maintains the wiki.
Getting started
bashnpx add-skill Astro-Han/karpathy-llm-wikiAfter installation, give the skill a URL, file, or pasted text to ingest a source. Ask questions like “What do I know about attention mechanisms?” to query the wiki. Run “Lint my wiki” to check for broken links, missing index entries, and stale cross-references.
License and pricing
Licensed under MIT. The project is open source with no stated pricing.
Based on the project's public README and repository metadata.