skills.rest
linaproai
linaproaiOFFICIAL
Explore all skills by this author →
WebsiteGitHub

karpathy-guidelines

Enforce behavioral guardrails for minimal, verified code changes.

147|28|Updated Apr 24, 2026
Search Tags:#code quality#code review#refactoring#software development#coding guidelines#llm best practices
One-click install
npx skills add https://github.com/linaproai/linapro --skill karpathy-guidelines-linaproai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: karpathy-guidelines
Source: https://github.com/linaproai/linapro/tree/main/.agents/skills/karpathy-guidelines
Command: npx skills add https://github.com/linaproai/linapro --skill karpathy-guidelines-linaproai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

LLMs frequently make common coding mistakes such as overcomplicating code, making unrequested changes to existing codebases, skipping verification of implementations, and hiding assumptions, which leads to wasted development time and buggy outputs. This Skill eliminates these issues by providing clear, actionable behavioral guidelines for all coding-related AI tasks.

Core Features & Use Cases

  • Pre-coding guardrails: Requires explicit surfacing of assumptions, tradeoffs, and ambiguities before writing any code to avoid misalignment with user intent.
  • Simplicity enforcement: Mandates minimal, request-aligned code with no unrequested features, abstractions, or speculative error handling to reduce technical debt.
  • Surgical editing rules: Ensures code changes only touch requested areas, match existing code style, and clean up only self-induced unused code to avoid unintended side effects.
  • Goal-driven verification: Requires defining clear, verifiable success criteria for all tasks to enable independent, accurate execution without constant clarification.
  • Use Case: Ideal for developers using AI assistants for code writing, pull request review, or legacy code refactoring to cut rework and ensure all AI-generated code changes are precise, minimal, and aligned with project requirements.

Quick Start

Use the karpathy-guidelines skill to review your latest AI-generated code refactor and flag any overcomplicated logic or unrequested style changes.

Frequently Asked Questions about karpathy-guidelines

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stop AI from making unrequested code changes during refactoring?▼

To stop unrequested code changes during refactoring, apply surgical editing guardrails that restrict modifications strictly to requested areas and mandate matching existing code style. This prevents unintended side effects and scope creep.

What is the best way to prevent LLMs from overcomplicating code generation?▼

Preventing LLMs from overcomplicating code generation requires enforcing simplicity guidelines that mandate minimal, request-aligned outputs. This restricts AI from adding unrequested features, speculative abstractions, or unnecessary error handling.

How do you define verifiable success criteria for AI coding tasks?▼

Defining verifiable success criteria for AI coding tasks involves setting explicit, goal-driven checkpoints before implementation. This enables independent execution and accurate verification without requiring constant user clarification.

Why does AI hide assumptions when writing software code?▼

AI hides assumptions when writing software code due to missing pre-coding guardrails. Forcing explicit surfacing of assumptions, tradeoffs, and ambiguities before writing code eliminates this issue and aligns outputs with user intent.

Can I use behavioral coding guidelines for pull request reviews across any framework?▼

Yes, you can use behavioral coding guidelines for pull request reviews across any programming language or framework. The guidelines apply universally to code writing, reviewing, and refactoring scenarios to flag overcomplicated logic.

Related Skills

Explore More Software Engineering Skills→
RugkeyProRugkeyPro

karpathy-guidelines

Provides behavioral guidelines for LLMs writing, reviewing, or refactoring code to reduce mistakes.

Community
Intermediate
chrislawcodeschrislawcodes

feature-spec

Generate structured feature specification documents in Markdown format.

Community
Advanced
0xRayAI0xRayAI

lint

Run node-based linting servers to enforce code quality standards.

Official
Intermediate
wangdaqian08wangdaqian08

karpathy-guidelines

Enforce minimal, surgical, and verifiable code changes with explicit tradeoffs.

Community
Basic