Yesterday I was working with an agent refactoring some code.
The agent applied the changes, praised itself for the nice code, and tried to commit. It failed to commit and then reset the changes. 😂
The next step was "Oops, I removed the changes, I will now recreate them". 😂
Why did it fail to commit? Because there was no local Git settings available to use.
The reason for that is that I run it inside a container. Call me paranoid, but I would not run an AI agent on my machine without any separation.
How do I separate agents from my host machine using containers? I wrote a bunch of wrappers to run them and Containerfiles to build them.
Everything is available at https://github.com/elwin013/agent-container
Kamil Banach