GameReverie
A Codex Skill for building and iterating games
Start new thread

GameReverie - A Codex Skill for building and iterating games

by
•
GameReverie is an open-source Codex Skill for taking a game from idea to first playable and through ongoing iteration. It separates design, implementation, and independent review, preserves project decisions across sessions, and turns playtest feedback into follow-up work. Guide development yourself or delegate a bounded goal. Developed with GPT-6 Astra, with a playable Snake demo and development records. Free and MIT-licensed.

Add a comment

Replies

Best
Maker
📌

Hi everyone — I’m the maker of GameReverie, an open-source Codex Skill for game development. I built and released it over three days with GPT-6 Astra for the Astra Challenge.

I wanted a workflow that wouldn’t stop at the first playable. I still need to play the game, figure out what feels off, discuss changes, and get them implemented and reviewed. GameReverie gives design, coordination, implementation, and review separate roles. Decisions and progress stay in project documents, so I can pick up the work in another session.

To try it out, I used GameReverie to build a Godot Snake game with one-way ramps, then added dash and whole-body jumps. Astra handled design, coordination, and technical decisions; Luna implemented tasks, and Sol independently reviewed the changes. That’s the model setup I used for the demo, not a requirement.

The reviewer caught a timing bug in the landing animation, which was fixed and reviewed again. My own playtesting led to changes in movement and ramp interactions. Those corrections are part of the demo too. The video is condensed from real sessions and gameplay.

GameReverie is free and MIT-licensed. The repo includes the Skill, a playable demo, and development records.

Where does AI tend to get stuck when you’re working on a game?

OpenAI
GPT-6 Astra Challenge
How did Astra change the scope or ambition of what you built?
Maker
The Astra Challenge prompted me to build and share GameReverie. Over three days, I worked with GPT-6 Astra to develop and release an MIT-licensed Codex Skill for game development. Working with Astra helped me pursue both a reusable workflow and a playable demonstration of it. The ambition was to support the full development cycle: discussing an idea, building the first playable version, reviewing changes, and incorporating human playtest feedback into further iterations. I then used GameReverie to develop a Snake game with one-way ramps, later adding dash and whole-body jumps. Astra handled design, coordination, and technical judgment; Luna implemented assigned tasks, and Sol independently reviewed them. Human playtesting guided revisions. The result includes the Skill, a playable demo, and development records showing how design decisions, review findings, and player feedback shaped the same project.