Karpathy-Inspired Claude Skills 🚀
SEO Keywords: Claude Skills, AI Coding Best Practices, Karpathy Guidelines, Claude Code Plugin, Cursor Rules, LLM Coding Pitfalls, Sustainable AI Engineering.
Check out my new project Multica — an open-source platform for running and managing coding agents with reusable Claude Skills. 🤖
Follow me on X: https://x.com/jiayuan_jy 🐦
🌟 Overview
A collection of modular, premium Claude Skills designed to drastically improve Claude Code behavior and reduce common LLM coding mistakes. These principles are derived from Andrej Karpathy's observations on AI coding pitfalls.
English | 简体中文 | हिन्दी | বাংলা | मराठी | తెలుగు | தமிழ் | ગુજરાતી | اردو | ಕನ್ನಡ
⚠️ The Problems
AI models often:
- ❌ Make wrong assumptions without checking.
- ❌ Overcomplicate code and APIs with bloated abstractions.
- ❌ Remove orthogonal code or comments they don't understand.
- ❌ Lack clear success criteria for complex tasks.
🛡️ The Solution: Premium Claude Skills
These modular Claude Skills directly address the issues above:
| Skill | Description | Addresses |
|---|
| 🧠 Think Before Coding | Forces explicit reasoning and surfaces assumptions. | Hidden confusion, wrong assumptions. |
| 🍃 Simplicity First | Minimal code that solves the problem. No speculation. | Overcomplication, bloated abstractions. |
| 🎯 Surgical Changes | Touch only what you must. Match existing style. | Orthogonal edits, style drift. |
| ✅ Goal-Driven Execution | Define success criteria and loop until verified. | Imperative vs Declarative gaps. |
📦 Modular Claude Skills in Detail
1. Think Before Coding
Don't assume. Don't hide confusion. Surface tradeoffs.
This skill ensures your AI partner explicitly reasons before touching a single line of code.
2. Simplicity First
Minimum code that solves the problem. Nothing speculative.
Combat the tendency toward overengineering. If 200 lines could be 50, rewrite it.
3. Surgical Changes
Touch only what you must. Clean up only your own mess.
Focus on precise edits. Match the project's existing style even if you'd do it differently.
4. Goal-Driven Execution
Define success criteria. Loop until verified.
Transform vague tasks into verifiable goals with specific success metrics.
🚀 Installation
Option A: Claude Code Plugin (Recommended)
Add the marketplace and install the Claude Skills:
/plugin marketplace add forrestchang/claude-skills-by-karpathy
/plugin install claude-skills-by-karpathy@karpathy-skills
Option B: CLAUDE.md (Per-Project)
curl -o CLAUDE.md https://raw.githubusercontent.com/forrestchang/claude-skills-by-karpathy/main/CLAUDE.md
🛠️ Tool Support
🤝 Contributing
Contributions are welcome! If you have a new Claude Skill or a refinement to existing ones, please open a PR.
📄 License
MIT © forrestchang
