Harness Hub
All open source

Open sourceMarkdown · ≈226k stars · See repository

Andrej Karpathy Skills

One CLAUDE.md that fixes common LLM coding pitfalls, based on Karpathy's notes.

multica-ai/andrej-karpathy-skills

Maintained by multica-ai

Coding Agents

A single CLAUDE.md file of guidelines, drawn from Andrej Karpathy's observations on LLM coding pitfalls, that improves how Claude Code behaves. You can install it as a Claude Code plugin from its marketplace or copy it in as a rules file, and it works in Cursor too.

View on GitHub

Pros & cons

Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.

Pros

  • DocsA single CLAUDE.md with four principles (Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution); it costs almost nothing to try.[1][2]
  • DocsIt targets agent failure modes taken from Karpathy's post: silent wrong assumptions, over-built abstractions, and edits outside the task.[1][2]
  • MetadataAbout 226k stars, though stars do not prove the rules get reused.[1]

Cons

  • User reportsKarpathy did not write or endorse the file, and this repo republishes the same content as the forrestchang repo its install instructions still point to.[1][2]
  • MetadataThe repository has no license file, so reuse terms are unclear.[1]
  • MetadataLast pushed in April 2026, and the README now opens by promoting the author's Multica platform.[1]
  • User reportsOne analyst predicts Claude's default behavior will absorb rules like these, which would limit their lasting value.[1]
Best for
Anyone who wants a quick, low-risk behavioral nudge against over-engineering in Claude Code or similar agents.
Watch out for
It is a prompt, not a tool; copy the four rules into your own AGENTS.md rather than depending on an unlicensed repo.

Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction

Neighbours on the workbench

MDX · ≈307k stars · MIT

superpowers

obra/superpowers

An agentic skills framework and software methodology that works.

Maintained by Jesse Vincent (obra)

4 pros · 4 cons

Shell · ≈306k stars · MIT

Matt Pocock Skills

mattpocock/skills

Small, composable agent skills for real engineering work.

Maintained by Matt Pocock (AI Hero)

4 pros · 4 cons

JavaScript · ≈292k stars · MIT

ECC

affaan-m/ECC

Agent harness performance optimization system: skills, instincts, memory and security for coding agents.

Maintained by Affaan Mustafa

4 pros · 4 cons

Connections

Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.

Maintained by · 1

  • multica-aiOrganisation

Built on · 1

Similar entries · 1

Sources

  1. GitHub repo

Verified Oct 29, 2026