agent0ai/agent-zero ? reverse-engineered prompt

Build me a personal AI assistant framework in Python. I'm looking for a general-purpose helper that isn't pre-programmed for specific tasks but can learn and grow as I use it. It should be able to use my computer to get things done, like running terminal commands, searching the web, and even writing its own code to solve problems. It needs to have a persistent memory so it can remember past solutions and facts.

A key feature should be multi-agent cooperation. I want the main agent to be able to create and delegate sub-tasks to other "subordinate" agents to help it solve more complex problems. It should also include a browser agent for web automation.

The whole system should be very customizable. I want to be able to change the agent's core behavior by editing simple text prompts, not by changing a bunch of code. Please include a web UI so I can interact with it, see its progress in real-time, and stop it if needed. Package the whole thing with Docker for easy setup.