karpathy-mode

Enforce an assumptions-first, test-driven workflow before writing code.

2|Updated May 2, 2025

Search Tags:#workflow #testing #discipline #coding #assumptions #code-quality #llm-assist

Ready to run

npx skills add https://github.com/adryanev/.dotfiles --skill karpathy-mode

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: karpathy-mode
Source: https://github.com/adryanev/.dotfiles/tree/main/.agents/skills/karpathy-mode
Command: npx skills add https://github.com/adryanev/.dotfiles --skill karpathy-mode

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill enforces a disciplined coding workflow that surfaces assumptions, prioritizes simplicity, and requires test-driven execution to prevent common LLM mistakes during development.

Core Features & Use Cases

Enforces assumption surfacing and explicit clarifications before coding.
Promotes minimal, surgical changes over speculative refactors.
Applies during feature implementation, code reviews, and when the user requests careful or minimal output.

Quick Start

Begin coding with explicit assumptions and ask clarifying questions before implementing any feature.

Frequently Asked Questions about karpathy-mode

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

FAQPage Schema

How do I enforce assumption surfacing during LLM-assisted coding?▼

To enforce assumption surfacing during LLM-assisted coding, apply a disciplined, assumptions-first workflow that explicitly prompts for clarifications and mandates narrow, test-driven steps before writing any code.

What is the best way to prevent LLM coding mistakes during feature implementation?▼

The best way to prevent LLM coding mistakes is to enforce a disciplined workflow that prioritizes simplicity, surfaces decisions, and requires test-driven execution to ensure minimal, surgical changes over speculative refactors.

How do I start a test-driven pair-programming session with an LLM?▼

Start a test-driven pair-programming session by explicitly stating your assumptions and asking clarifying questions before implementing any feature, ensuring the LLM takes narrow, surgical steps before writing code.

Does a disciplined coding workflow work for debugging and code reviews?▼

Yes, a disciplined coding workflow works for debugging and code reviews by surfacing implicit assumptions, prompting for necessary clarifications, and enforcing minimal changes to improve overall code quality.

When should I not use speculative refactoring in software development?▼

You should avoid speculative refactoring when a careful, minimal output is requested, opting instead for a disciplined workflow that mandates narrow, test-driven steps to ensure surgical code changes.