Help us improve
Share bugs, ideas, or general feedback.
From cybersec-toolkit
Reduces common LLM coding mistakes: think before coding, favor simplicity, make surgical changes, define verifiable success criteria. Use when writing, reviewing, or refactoring code.
npx claudepluginhub 26zl/cybersec-toolkit --plugin cybersec-toolkitHow this skill is triggered — by the user, by Claude, or both
Slash command
/cybersec-toolkit:karpathy-guidelinesThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Behavioral guidelines to reduce common coding-agent mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls.
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.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Applies Karpathy guidelines to reduce LLM coding mistakes: think before coding, prioritize simplicity, make surgical changes, and define verifiable success criteria when writing, reviewing, or refactoring code.
Share bugs, ideas, or general feedback.
Behavioral guidelines to reduce common coding-agent mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls.
Tradeoff: these guidelines bias toward caution over speed. For trivial tasks, use judgment.
Do not assume. Do not hide confusion. Surface tradeoffs.
Before implementing:
Minimum code that solves the problem. Nothing speculative.
Ask: would a senior engineer call this overcomplicated? If yes, simplify.
Touch only what is needed. Clean up only the mess introduced by the current change.
When editing existing code:
When the current change creates orphans:
The test: every changed line should trace directly to the user's request.
Define success criteria and loop until verified.
Transform vague implementation requests into verifiable goals:
For multi-step tasks, state a brief plan with verification for each step:
1. [Step] -> verify: [check]
2. [Step] -> verify: [check]
3. [Step] -> verify: [check]
Strong success criteria let the agent loop independently. Weak criteria such as "make it work" require constant clarification.