andrej-karpathy-skills
multica-ai/andrej-karpathy-skills
A behavioral guidelines framework for coding assistants derived from Andrej Karpathy's observations on common LLM pitfalls, emphasizing simplicity, surgical changes, and goal-driven execution.
201.9k 20.7kNo licenseUpdated 2026-04-20
69
GoodOverview
This skill implements four core principles—Think Before Coding, Simplicity First, Surgical Changes, and Goal-Driven Execution—designed to curb common AI coding mistakes such as overcomplication, unauthorized refactoring, hidden assumptions, and lack of verification. Delivered primarily as a CLAUDE.md or cursor rule file, it restrains the agent's tendency to generate bloated abstractions and ensures changes are minimal, precise, and test-verified.
Capabilities
- ▸Enforces strict adherence to minimal, non-speculative coding practices
- ▸Guides agents to surface uncertainties and trade-offs before implementing
- ▸Restricts code diffs strictly to the user's explicit request (surgical changes)
- ▸Transforms vague tasks into verifiable, test-driven execution loops
Best for
Configuring Claude Code or Cursor agents to write cleaner, less over-engineered code., Reviewing and refactoring codebases with an explicit emphasis on minimal modifications and strict adherence to user requests., Preventing LLMs from making unverified assumptions or modifying adjacent, unrelated code files.
Works with
Claude CodeCursorCodexGitHub Copilot
Evoa Score breakdown= Σ (score × weight)
69Task usefulness20%77 → +15.4
実タスクにどれだけ役立つか（機能の豊富さ・用途の明確さ）。 — AIによるcapabilities/use-cases解析
Code quality15%95 → +14.3
実装・指示の品質。 — AIによるSKILL.md/README解析
Maintenance15%70 → +10.5
リポジトリがどれだけ活発に保守されているか。 — GitHub 最終push日時の新しさ
Documentation12%42 → +5.0
ドキュメントの充実度・分かりやすさ。 — README/独自要約の情報量
Security15%100 → +15.0
危険・不審な挙動が無いか。 — AIによるセキュリティレビュー
Originality10%85 → +8.5
ありふれたラッパーではない独自性。 — AIによる独自性判定
Popularity8%100 → +8.0
コミュニティの採用度。 — GitHub Stars/Forks（対数スケール）
Compatibility5%100 → +5.0
対応AIエージェントの広さ。 — AIによる対応エージェント判定
Weighted total81.7 / 100
ライセンス不明/制限あり（Red）のSkillは総合スコアに0.85倍の補正を適用します。 ランキングはこのScoreのみで決まり、広告で変わりません。 算出方法の詳細 →
Security considerations
This skill consists entirely of behavioral guidelines and instructions for coding assistants; it contains no executable code or network access capabilities, presenting zero direct security risk.
Categories
Summary and analysis are original content generated by AI Skills Rank. The skill's source text is not reproduced here — view it on the linked repository.