lone dragon
We have an agent-controlled upgrade flow where Docker Compose and Git are available only on the Windows host, while the application CLI runs inside a Docker container via docker exec.
We are choosing between:
- a Python driver inside the container that somehow asks the host to run Compose,
- direct agent orchestration of every command through host exec,
- a dedicated host-side PowerShell runner invoked by the agent.
Is option 3 the recommended architecture?
Specific validation needed:
- Can a PowerShell script safely run ~10
docker exec <container> <cli>commands in parallel? - Can it write outputs and a manifest atomically to a host/container shared workspace?
- Is there any reason to prefer a container-side Python driver with host command hand-off over a host-side deterministic runner?
sinful trout