TransformerOptimus/SuperAGI ? reverse-engineered prompt

I want to build a complete framework for creating and running my own autonomous AI agents. It should be a full-stack application with a nice graphical user interface where I can configure and launch new agents, give them specific goals, and then monitor their progress.

The agents need to be able to perform actual tasks, so they should have access to a library of tools for things like browsing the web, posting to Twitter, reading files, or even writing code. The system should make it easy to add new tools, too.

From the UI, I'll need an action console to interact with an agent while it's running, maybe to give it feedback or approve a critical step. The agents should also have memory to learn from their actions. Finally, please include performance telemetry so I can see how well they're doing and track my LLM token usage to control costs.