Andrej Karpathy Skills is a single-file set of behavioral guidelines that address the most common LLM coding pitfalls — derived from Andrej Karpathy’s observations on how language models go wrong when writing code. Install it as a Claude Code plugin, a Cursor project rule, or a per-projectDocumentation Index
Fetch the complete documentation index at: https://mintlify.com/multica-ai/andrej-karpathy-skills/llms.txt
Use this file to discover all available pages before exploring further.
CLAUDE.md file.
Get started
Install the guidelines as a Claude Code plugin or drop-in CLAUDE.md file in under 2 minutes.
The four principles
Explore each principle — Think Before Coding, Simplicity First, Surgical Changes, and Goal-Driven Execution.
Real-world examples
Side-by-side examples of what LLMs do wrong and how the guidelines correct the behavior.
Cursor setup
Use the same guidelines in Cursor via committed project rules — no extra configuration needed.
The problem these guidelines solve
From Andrej Karpathy:“The models make wrong assumptions on your behalf and just run along with them without checking. They don’t manage their confusion, don’t seek clarifications, don’t surface inconsistencies, don’t present tradeoffs, don’t push back when they should.”
“They really like to overcomplicate code and APIs, bloat abstractions, don’t clean up dead code… implement a bloated construction over 1000 lines when 100 would do.”These four principles directly address each failure mode.
The four principles at a glance
Think Before Coding
State assumptions explicitly. Surface tradeoffs. Ask rather than guess.
Simplicity First
Write the minimum code that solves the problem. Nothing speculative.
Surgical Changes
Touch only what you must. Match existing style. Clean up only your own mess.
Goal-Driven Execution
Define verifiable success criteria. Transform tasks into testable goals.
Quick install
Prefer a per-project file? See Installation for the
CLAUDE.md drop-in option and how to merge with existing project instructions.