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.
The 'karpathy-guidelines' skill provides a set of behavioral guidelines aimed at reducing common coding mistakes associated with large language models (LLMs). By emphasizing caution over speed, this skill helps developers avoid overcomplicated solutions and encourages them to make surgical changes to code. It addresses the frequent pitfalls observed in LLM coding practices, ensuring that developers can write, review, and refactor code more effectively and efficiently. The guidelines promote clarity in assumptions, simplicity in code, and a goal-driven approach to execution, ultimately leading to better coding practices and outcomes.
You can integrate the guidelines by following the outlined principles during writing, reviewing, or refactoring code.
Yes, the principles can be applied across various programming languages as they focus on coding practices rather than specific language features.
The guidelines may not cover every specific scenario, and developers should use their judgment for trivial tasks.
Absolutely, they promote clear communication and shared understanding, which are beneficial in collaborative settings.
These guidelines are particularly useful for software engineers, code reviewers, and anyone involved in coding tasks with LLMs.
Quick Setup:
.claude/skills/Repository
forrestchang/andrej-karpathy-skills