godogen — Claude Code Skills That Build Complete Godot Games From Text
godogen is an open-source project that uses Claude Code skills to transform natural language game descriptions into fully functional Godot 4 projects. Trending on Hacker News with 141 points, it orchestrates an AI pipeline that designs architecture, generates 2D art (via Gemini), converts images to 3D models (via Tripo3D), writes all GDScript code, captures in-game screenshots, and automatically fixes visual problems.
The system uses two specialized Claude Code skills — one for planning, one for execution — with each task running in a fresh context to maintain focus. It includes custom-built language references and lazy-loaded API documentation for all 850+ Godot classes, compensating for LLMs' thin training data on GDScript.
A visual QA loop captures and analyzes screenshots from the running Godot engine to identify and fix rendering issues autonomously. The entire pipeline runs on commodity hardware — any PC with Godot 4 and Claude Code.
This represents a new frontier for coding agents: not just writing code, but orchestrating complete creative pipelines that span architecture, art generation, code generation, and visual testing.
GitHub: https://github.com/htdt/godogen
← Back to all articles
The system uses two specialized Claude Code skills — one for planning, one for execution — with each task running in a fresh context to maintain focus. It includes custom-built language references and lazy-loaded API documentation for all 850+ Godot classes, compensating for LLMs' thin training data on GDScript.
A visual QA loop captures and analyzes screenshots from the running Godot engine to identify and fix rendering issues autonomously. The entire pipeline runs on commodity hardware — any PC with Godot 4 and Claude Code.
This represents a new frontier for coding agents: not just writing code, but orchestrating complete creative pipelines that span architecture, art generation, code generation, and visual testing.
GitHub: https://github.com/htdt/godogen