WingedGuardian/GENesis-AGI ? reverse-engineered prompt

Build me a personal AI system that can run on my own Linux machine and keep working between sessions, not just answer one prompt and forget everything.

I want it to feel like a persistent helper that remembers my projects, learns from what works and what fails, and can start taking useful actions on its own when it has enough confidence. It should have a clear dashboard so I can see what it is doing, what is healthy, and what it is planning. It should also support a solid memory system so it can store important facts, recall them later, and improve over time from past outcomes.

Please make it practical and runnable, with a setup flow for a local environment, config files for secrets and tools, and enough structure that it can use browser automation, research, monitoring, and other real world actions. If you need to check current docs for anything like Claude Code or Playwright, go ahead and look them up online. Keep it robust, testable, and designed for long running autonomous use, not just a demo.