github.devotd.
NewAI agents

rehan-remade/universal-modder — a game modding tool for agents

rehan-remade/universal-modder provides skills and a knowledge base that help AI agents mod the PC games they have access to.

Repositoryrehan-remade/universal-modder ↗

Point Claude at any game. Skills, tools and the fal MCP that let Claude Code mod almost any PC game you own: recon, reverse engineering, fal-generated art/3D/audio, in-game testing, showcase videos.

stars
★ 8.3K
Forks
899
Language
Python
License
MIT
Read original source ↗View on GitHub ↗

rehan-remade/universal-modder is a set of skills and a CLI package that help you mod PC games you have access to in Claude Code, Codex, Cursor, and similar tools. It provides a loop of finding a game, identifying its engine, analyzing the code, building a mod, and then testing it in the running game. The latest release version is not listed in FACTS.

Its distinguishing feature is a knowledge base that reuses per-game investigation records in later work. Under knowledge/ there are 164 field notes covering 98 games, holding reverse-engineering results such as engine identification, packet structures, and decompilation paths. There are 12 skills, including game-recon, reverse-engineering, and asset-pipeline. The repository contains 54 Python sources and 7 test files, and it also supports art, 3D, and sound generation through the fal MCP. Installation is documented through per-agent plugin commands or through npx skills add and the um CLI.

That said, the results material in the README leans mainly on demo GIFs, and the repository has no quantitative evaluation backing it. Creating derived assets requires a fal API key, but when you have no key you can use fal's OAuth login or a local ComfyUI. So the modding procedure and the way knowledge is shared can be checked, but the quality of the results is hard to judge from the published material alone.