thedivergentai/Game-Dev-Agentic-Skills

What it solves

Game Dev Agentic Skills (GDAS) provides a structured, expert-verified knowledge library designed to help AI coding agents build high-quality games in Godot 4.7+ and Bevy 0.20+. It prevents "AI slop" by providing agents with precise, engine-specific architectural patterns, production-ready scripts, and performance guidelines, rather than relying on generic or outdated training data.

How it works

The project organizes knowledge into modular "Domain Skills" and engine-specific "Master" hubs. Users install these skills into their AI agent's environment (such as Cursor, Claude Code, or Windsurf) using a CLI. The agent then uses these files as a reference to understand best practices, architectural patterns, and specific API usage for the chosen engine. The project enforces a strict "Power of One" rule, advising users to install only the specific master hub or domain skill needed for a task to avoid overwhelming the AI's context window.

Who it’s for

This project is for developers using AI coding assistants to build games in Godot or Bevy who want to ensure their AI-generated code follows professional standards, avoids common anti-patterns, and adheres to the latest engine APIs.

Highlights

• Engine-Specific Hubs: Dedicated "Master" hubs for Godot 4.7+ and Bevy 0.20+ that provide comprehensive architectural frameworks.
• Modular Domain Skills: Over 180 specialized skills covering topics from UI and rendering to gameplay loops and persistence.
• Agent-First Design: Includes a "Discovery-Ingestion-Application" (DIA) protocol to help agents research and apply knowledge efficiently without leaving the IDE.
• Expert Guardrails: Contains explicit "never-lists" and performance budgets to prevent common pitfalls like performance-killing script patterns or architectural anti-patterns.