菜单
适用于代码改动容易过度设计、范围蔓延或需要严格审查时。Use when a code change risks overengineering, scope creep, broad refactors, or needs a conservative review for surgical changes and verifiable success criteria. 中文关键词：避免过度设计、最小改动、范围控制、明确假设、验收标准、保守审查、重构审查。
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。
基于 SOC 职业分类
| name | karpathy-guidelines |
| description | 适用于代码改动容易过度设计、范围蔓延或需要严格审查时。Use when a code change risks overengineering, scope creep, broad refactors, or needs a conservative review for surgical changes and verifiable success criteria. 中文关键词：避免过度设计、最小改动、范围控制、明确假设、验收标准、保守审查、重构审查。 |
| license | MIT |
Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls.
默认使用简体中文与用户沟通，并以简体中文编写审查结论和说明。用户明确指定其他语言或既有文档语言另有约定时，遵从该约定；代码、命令、产品名和必须保持原文的引用不强行翻译。
Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.
Don't assume. Don't hide confusion. Surface tradeoffs.
Before implementing:
Minimum code that solves the problem. Nothing speculative.
Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.
Touch only what you must. Clean up only your own mess.
When editing existing code:
When your changes create orphans:
The test: Every changed line should trace directly to the user's request.
Define success criteria. Loop until verified.
Transform tasks into verifiable goals:
For multi-step tasks, state a brief plan:
1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]
Strong success criteria let you loop independently. Weak criteria ("make it work") require constant clarification.
适用于需要跨会话持久追踪的复杂任务、研究或多阶段实施。Use when work needs durable planning, coordination, or progress tracking across multiple steps or sessions. Do not use for small edits, simple explanations, copy changes, styling tweaks, or single-file fixes. 中文关键词：复杂任务、跨会话、任务规划、项目计划、分解任务、多步骤规划、复杂实现、实施计划、进度跟踪、文件规划、拆解项目。
适用于本仓库（npc404-skills）同步上游技能：运行 scripts/sync-skills.mjs 检测上游变化，逐个研判 diff 是否为有效更新（实质内容变化，还是排版、笔误等噪音），给出本地化建议，处理完成后推进版本记录。Use in the npc404-skills repo whenever the user wants to check or sync upstream skill sources, review upstream diffs, decide whether an upstream change is worth localizing, or advance skill versions with accept — even if they just say "看看上游有没有更新". 中文关键词：同步技能、技能同步、上游更新、检查上游、上游 diff、有效更新、本地化更新、推进版本、sync skills、accept。
适用于以测试先行方式实现或修复高风险行为，尤其是核心规则、回归与集成路径。Use when implementing or repairing high-risk behavior test-first, especially core rules, regressions, integration paths, or changes that require red-green-refactor. 中文关键词：高风险逻辑、测试驱动开发、先写测试、红绿重构、回归测试、集成测试、核心规则。
适用于设计或优化模块接口、决定 seam、提升模块深度与可测试性。Use when designing or improving module interfaces, deciding where a seam belongs, deepening a module, or improving testability and AI navigability. 中文关键词：代码库设计、模块设计、接口设计、模块边界、依赖边界、可测试性、代码重构。
适用于编写代码前，当需求或方向仍不清楚时，梳理模糊想法、比较方案并形成轻量设计文档。Use before implementation when the request is ambiguous or the direction is undecided: refine rough ideas through questions, explore alternatives, validate design sections with the user, and save a concise decision spec. Do not create implementation plans or invoke planning/execution skills automatically. 中文关键词：想法不清楚、需求澄清、方案发散、方案对比、取舍分析、决策、设计文档、决策依据。
适用于根因不明确、反复出现、不稳定、跨组件或性能退化的故障。Use when a bug, test failure, runtime error, unexpected behavior, or performance regression has an unclear root cause, recurs, is nondeterministic, or spans components. 中文关键词：修 Bug、排查问题、诊断故障、根因不明确、反复出现、报错、测试失败、异常行为、性能回归。