| name | container-git-environment |
| description | Use for any agent running inside a Docker container that interacts with the project's git repository. Explains the bare repo clone model, branch naming, and how work is persisted. |
Container Git Environment
Containers interact with a bare git repository on the host. This skill describes the git data flow.
Branch Model
Each container agent works on its own branch:
docker/{project-id}/{agent-name}
The seed branch is docker/{project-id}/current-root, which is synced from the exterior (host) repo. Agent branches fork from docker/{project-id}/current-root on first launch.
Clone
The container's entrypoint clones from the bare repo at startup. Your working directory is a normal git checkout of your agent's branch.
How Work Is Persisted
Your access scope determines how commits reach the bare repo:
-
Write agents: Every git commit is automatically pushed to docker/{project-id}/{agent-name} on the bare repo. You never need to run git push — a PostToolUse hook handles it. Branch switching is blocked — you stay on your assigned branch.
-
Build intercept agents (UE): The build hook commits and pushes your changes before routing the build to the host. Manual git push is blocked. Branch switching is blocked.
-
Read-only agents: Your workspace is ephemeral. No changes persist. You can freely switch branches and navigate git history.
Reading Other Branches
All agents can read any branch without switching:
git show docker/{project-id}/agent-2:Source/MyFile.cpp
git log docker/{project-id}/current-root --oneline -10
git diff HEAD..docker/{project-id}/current-root -- Source/
Visibility
- Your branch is visible to the coordination server and other agents that fetch it.
- The bare repo is persistent — created once by
setup.sh, never recreated on launch.
