using-git-worktrees

Git worktrees allow a single repository to host multiple active working directories, enabling parallel branch checkouts so concurrent tasks do not conflict over the same checkout. This avoids repeatedly cloning the repository.

Isolating parallel feature development is best achieved by creating sibling worktrees for separate branches. This allows agents or developers to work in parallel without stomping on each other's changes within the same shared object database.

You create a sibling worktree by checking out the target branch into a dedicated directory, such as adding a worktree at a sibling path for the feature branch. This ensures the agent works only within that isolated directory.

Using git worktrees does not require duplicating the repository object database. Worktrees preserve a shared object database while providing separate writable surfaces for each concurrent task or branch.

Git worktrees are applicable for multi-agent orchestration, hotfixes, and code reviews where separate writable surfaces are required. They provide operational guidance to direct subagents into the correct worktree and keep writable scope narrow.

You can manage long-lived experiments by inspecting, entering, and removing git worktrees. This operational guidance helps maintain simultaneous feature development streams and cleans up isolated working directories when they are no longer needed.