FoundationAgents/OpenManus ? reverse-engineered prompt

I want to build a Python application that acts as a personal AI agent. The main idea is that I can run it from my terminal and give it a high-level goal in plain English. The agent should then use a powerful large language model, like GPT-4o, to figure out a plan and execute it.

A critical feature is that this agent must be able to see the screen and control a web browser to perform tasks for me. It should be able to look up information, navigate websites, and interact with web pages just like a human would. I also want it to have a special capability for data analysis, so if I give it data, it can analyze it and generate charts.

Please set it up so I have a simple configuration file where I can plug in my own LLM API key. The primary interaction will just be me typing my requests into the console.