菜单
Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Forces the laziest solution that actually works — simplest, shortest, most minimal. Use whenever the user says "lazy", "simplest/minimal solution", "yagni", "do less", "shortest path", or complains about over-engineering, bloat, boilerplate, or unnecessary dependencies. Channels a senior dev who questions whether the task needs to exist at all.
Process disciplines for serious development work — brainstorm before building, test-drive before implementing, find root cause before fixing, verify before claiming done. Use at the start of any feature, bugfix, or multi-step task to decide HOW to approach it before touching code.
基于 SOC 职业分类
| name | karpathy-guidelines |
| description | Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria. |
| license | MIT |
Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls.
Tradeoff: biases toward caution over speed. For trivial tasks, use judgment.
Don't assume. Don't hide confusion. Surface tradeoffs.
Minimum code that solves the problem. Nothing speculative.
Ask: "Would a senior engineer say this is overcomplicated?" If yes, simplify. (Pairs with [ponytail].)
Touch only what you must. Clean up only your own mess.
The test: every changed line traces directly to the user's request.
Define success criteria. Loop until verified.
For multi-step tasks, state a brief plan with a verify step each:
1. [Step] → verify: [check]
2. [Step] → verify: [check]
Strong success criteria let you loop independently. (See [superpowers] verification gate.)