Andrej Karpathy Skills — Claude Code 가이드라인
요약
🔗 원본: GitHub: multica-ai/andrej-karpathy-skills
Andrej Karpathy의 LLM 코딩 문제점 관찰에서 비롯된 4가지 원칙을 담은 CLAUDE.md 파일. Claude Code의 행동을 개선하기 위해 설계됨.
주요 내용
4가지 원칙
| 원칙 | 해결하는 문제 |
|---|---|
| Think Before Coding | 잘못된 가정, 숨겨진 혼란, 트레이드오프 누락 |
| Simplicity First | 과도한 엔지니어링, 부풀려진 추상화 |
| Surgical Changes | 관련 없는 코드 수정, 건드리면 안 되는 코드 변경 |
| Goal-Driven Execution | 테스트 우선, 검증 가능한 성공 기준 |
설치 방법
- Claude Code Plugin:
/plugin install andrej-karpathy-skills@karpathy-skills - CLAUDE.md 직접 추가:
curl -o CLAUDE.md https://raw.githubusercontent.com/forrestchang/andrej-karpathy-skills/main/CLAUDE.md
연결
- [[Andrej Karpathy]] — 원저자
- [[Think Before Coding]] — 4원칙 중 첫 번째
- [[Simplicity First]] — 4원칙 중 두 번째
- [[Surgical Changes]] — 4원칙 중 세 번째
- [[Goal-Driven Execution]] — 4원칙 중 네 번째
인용
“The models make wrong assumptions on your behalf and just run along with them without checking.” — Andrej Karpathy
“LLMs are exceptionally good at looping until they meet specific goals… Don’t tell it what to do, give it success criteria and watch it go.” — Andrej Karpathy
원문 (요약)
# Karpathy-Inspired Claude Code Guidelines
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
## The Problems
- LLMs make wrong assumptions without checking
- They overcomplicate code, bloat abstractions, don't clean up dead code
- They change/remove code they don't understand as side effects
## The Solution: Four Principles
1. Think Before Coding — State assumptions, surface tradeoffs, push back
2. Simplicity First — Minimum code, nothing speculative
3. Surgical Changes — Touch only what you must
4. Goal-Driven Execution — Define success criteria, loop until verified
## Install
- Option A: Claude Code Plugin (/plugin install andrej-karpathy-skills@karpathy-skills)
- Option B: CLAUDE.md (curl -o CLAUDE.md https://raw.githubusercontent.com/.../main/CLAUDE.md)