OPPO-Mente-Lab/X-OmniClaw — reverse-engineered prompt

Build me an Android app that works like a personal phone agent on a real device. I want to type or speak a task, then have the app look at the current screen, understand what’s happening, decide the next step, and carry it out with normal phone actions like tapping, swiping, typing, taking screenshots, and switching apps.

It should feel like an assistant that can handle tasks across different apps, not just chat. It should observe the result after each action, keep going until the task is done, and stop safely if it gets stuck or the user cancels. Please include basic memory so it can use previous context, plus a simple way to schedule repeated tasks by time, day, or interval.

Make the interface simple, with a text box, voice input, task status, logs of what it did, and start or stop controls. Use Kotlin for Android and target modern Android devices, but keep support for Android 8 and newer if practical. Look up current Android docs online if needed.