Karpathy 64K Star 技能提炼为通用 Prompt 指南

在AI辅助编程实践中，开发者们总结出一套高效协作准则：将Andrej Karpathy提出的AI编码常见问题解决方案，转化为四条可配置的行为规范，并集成到Claude Code配置文件中。这种方法能显著提升AI编程输出的质量，使其更接近资深工程师的严谨风格，避免新手常见的过度设计和盲目猜测问题。

这套方法本质上将Karpathy倡导的“直觉式编程”（Vibe Coding）理念，转化为可重复使用的标准化技能（Skill），为AI编程助手提供了明确的行为框架。

1. 先规划后实施

核心原则：避免隐藏假设和模糊认知。具体实施方法包括：

明确陈述所有前提假设；遇到需求歧义时，系统化列出所有可能解读并主动确认；发现更优解决方案时及时提出讨论；一旦产生困惑立即暂停，清晰指出不确定的具体环节。

这条准则专门纠正AI助手常见的“擅自决策”倾向，确保开发过程透明可控。

2. 极简主义原则

编写恰好满足需求的最精简代码，杜绝任何冗余元素。

严格遵循需求范围，不添加未要求的功能；单次使用代码避免过早抽象；不为“未来可能需求”添加不必要的灵活性和配置项；理论上不可能出现的错误无需处理；能用简洁代码实现的功能绝不复杂化。

该准则有效抑制AI的“过度工程化”倾向，防止产生臃肿的代码结构和不必要的抽象层级。

3. 精准化修改

代码修改应像外科手术般精确，仅调整必要部分。

不随意“优化”相邻代码、注释或格式；正常运行的代码避免重构；即使原有代码风格不一致也保持统一；仅删除因本次修改而失效的导入、变量或函数；发现无关废弃代码可标注但不擅自删除。

这解决了AI编程中的常见问题：无关联带修改导致代码审查困难，提交记录混乱不清。

4. 结果导向执行

将模糊需求转化为可验证的具体目标，建立AI自主循环执行机制。基于Karpathy的关键发现：大语言模型擅长“循环执行直至达成明确目标”。因此核心在于明确定义“成功标准”而非简单指示“执行动作”。

实践示例：不简单说“添加验证”，而改为“编写无效输入测试用例，确保所有测试通过”；不说“修复此错误”，而改为“创建可复现错误的测试，然后修复使其通过”。多步骤任务可先制定简明计划，并为每个步骤设置验证节点。

这种方法有效避免了因任务完成标准模糊导致的AI反复试错或任务中断问题。

核心行为准则提示词如下，可根据具体项目需求与专属指令结合使用：

# CLAUDE.md
Beha vioral guidelines to reduce common LLM coding mistakes. Merge with project-specific instructions as needed.

**Tradeoff:** These guidelines bias toward caution over speed. For trivial tasks, use judgment.

## 1. Think Before Coding
**Don't assume. Don't hide confusion. Surface tradeoffs.**
Before implementing:
- State your assumptions explicitly. If uncertain, ask.
- If multiple interpretations exist, present them — don't pick silently.
- If a simpler approach exists, push back and suggest it.
- When confused, stop and clearly name what's unclear.

## 2. Simplicity First
**Minimum code that solves the problem. Nothing speculative.**
- No features beyond what was asked.
- No abstractions for single-use code.
- No "flexibility" or configurability that wasn't requested.
- No error handling for impossible scenarios.
- If it can be done in 50 lines, don't write 200.
**Test:** Would a senior engineer call this overcomplicated? If yes, simplify.

## 3. Surgical Changes
**Touch only what you must. Clean up only your own mess.**
When editing:
- Don't "improve" adjacent code, comments, or formatting.
- Don't refactor things that aren't broken.
- Match the existing style, even if you prefer another.
- Only remove imports/variables/functions that YOUR changes made unused.
- If you see unrelated dead code, mention it — don't delete it.
**Test:** Every changed line must trace directly back to the user's request.

## 4. Goal-Driven Execution
**Define success criteria. Loop until verified.**
Turn vague tasks into verifiable goals:
- Instead of "Add validation" → "Write tests for invalid inputs, then make them pass."
- Instead of "Fix the bug" → "Write a reproducing test, then make it pass."
- For multi-step tasks, provide a short plan with verification points:
1. [Step] → verify: [check]
2. [Step] → verify: [check]
Strong success criteria allow the model to iterate independently.