Overview

This repository offers a foundational CLAUDE.md file designed to significantly enhance the coding behavior of Claude. Inspired by Andrej Karpathy's critical observations on Large Language Model (LLM) coding pitfalls, these guidelines provide a structured approach to mitigate common issues like unwarranted assumptions, code bloat, and unintended side effects.

Key Features

• Four Core Principles: Directly addresses LLM coding weaknesses with actionable principles: "Think Before Coding," "Simplicity First," "Surgical Changes," and "Goal-Driven Execution."
• Mitigates LLM Pitfalls: Tackles issues such as silent assumption-making, overcomplication, tangential code modifications, and confusion management.
• Promotes Explicit Reasoning: Encourages Claude to state assumptions, present tradeoffs, and ask clarifying questions rather than proceeding with potentially incorrect interpretations.
• Enforces Code Simplicity: Drives the generation of minimal, effective code, avoiding unnecessary abstractions or speculative features.
• Ensures Precise Edits: Guides Claude to make only the necessary changes, preventing accidental alterations to unrelated code or comments.
• Facilitates Verifiable Success: Transforms imperative tasks into declarative goals with clear success criteria, enabling more autonomous and reliable execution.
• Flexible Integration: Easily installable as a Claude Code plugin for universal use or embeddable directly into a project's CLAUDE.md file.

Typical Use Cases

• Improving LLM Code Generation: For developers seeking to obtain more accurate, efficient, and maintainable code from Claude.
• Reducing LLM-Induced Errors: Ideal for projects where the cost of LLM-generated code errors (e.g., debugging, refactoring) outweighs the initial development speed.
• Establishing Coding Standards for LLMs: A practical tool for teams to enforce consistent, high-quality coding practices when leveraging AI assistants.
• Guiding AI Pair Programming: Enhances the collaborative coding experience by ensuring the AI partner acts as a rigorous, focused contributor.
• Training and Fine-tuning LLM Behavior: Serves as a concrete example of desired coding patterns and rationales for AI models.