[AI Usage] A tutorial, as we progress towards the 0.8.5 version

[mental-os]

You can read more about the 0.8.5 milestone in issue #82

Inspired by @masterofmagic999 's question #89 , I've reflected of a way of synchronizing modern AI powered IDE's across contributors and environments.

As you already know from README.md 's disclaimer, Aurora OS.js is using AI agents assisting in documentation, GitHub integrations, bug testing, and roadmap tracking - yes, I hate documentation and linting to keep the codebase on the standard, especially when I have an idea and I want to make it NOW and fix it later.
I know many of you feel the same from time to time. The truth is, however, that AI is not yet at the level of creating without human help (and frankly, in a collaborative workspace, it shouldn't). This doesn't stop us to leverage its power in this ambitious project, tho - it would be a bit weird to bother us since we are some passionate cyberpunks building a hacking simulator, if you ask me.

So, ✨ how do you make good use of AI without it fucking everything up ✨?
In the latest commits you will see that I've started to take AI-usage into consideration more and more.
This gave us .agent (nightly) folder with different helpers trying to standardize the models and repository knowledge.
Out of the box it is fully compatible with Google Antigravity and Google Gemini 3 Pro (high and fast) - just clone the repository and Antigravity will know what to do with this folder.

It is still a work in progress thing that needs further enhancing but it is a starting point:

• context.md is a bird-eye view of the repository with the intended purpose of feeding most relevant and critical information. You will see specific instructions and rules to follow for the code we wrote and for the systems we built.
• in this order, update-translations.md > update-context.md > prepare-github.md, are workflows (invoked by /[name]) ment to standardize the pain I have everytime I commit my work. They handle the feeding/checking/updating of en.ts translation, the documenting of the session, and, of course, the linting, testing, building, and wrapping of a GitHub pull request (you will still have to check everything up).

All of these, and the repository itself, have a public correspondent for various online tools such as MCP servers or AI based forensics. Because no open-source project is truly open-source if we gatekeep the tricks.

Speaking of MCP servers, you can check GitMCP.io to quickly setup you favorite IDE to work with Aurora OS.js codebase.

That's it, haha!
Thank you for coming to my TED talk ^_^

Towards limitless creation, peeps.
#stayRebel

[mental-os]

Oh, and don't forget to always tell the AI to not make any mistakes.