Feature-Request: Isolate agents in containers

Hello!

I'm trying to use agents with spec-driven development methods. During that I've found that it would probably be better to run commands executed by an agent inside of an isolated container environment. My thought on this was that every instance of such a container with every agent should probably work on it's own working copy of the repository.

A few weeks later I've discovered this project and I would like to ask you to provide such a feature to run the agents in containers. I would love to hear your feedback.

Best,
Dennis

The issue names no files, tests, entry points, or container runtime. Start by locating how agents execute commands and how each agent gets a repository workspace; done requires an agreed approach for isolated containers and separate working copies, with coverage for the resulting behavior.

Written by the indexing model from the issue text.