kotmasha/variable_graph_MAS ? reverse-engineered prompt

Build me a Python simulation for multi agent navigation based on the plug and play cooperative navigation idea from the README.

I want to be able to run one main script with a YAML config file and a simulation time, then watch several agents move toward their goals while staying connected to each other and avoiding obstacles. The app should support different obstacle worlds, like circular, star shaped, and rectangular obstacles, and it should let me switch between a more aggressive controller and a more minimal effort one.

Please include clear configuration for things like agent names, leader targets, safe distance, communication range, and cooperation strength. I also want plots or animations that show the trajectories and the communication graph over time, so I can tell when the network stays connected.

If anything is unclear from the paper or README, look up the current docs online and make sensible defaults so the example runs out of the box.