Create Worktree

create-worktree

Invocation input

$ARGUMENTS below means the arguments supplied with the user's explicit invocation. In inherited Context blocks, run each ! command to collect the named value; those expressions are not expanded automatically in a skill.

Context

• Current branch: !git branch --show-current
• Current git status: !git status --short
• Existing worktrees: !git worktree list
• Recent commits: !git log --oneline -10

Your task

Create or resume a worktree for the work at hand and continue in it. $ARGUMENTS may include a Baseline Pointer: an exact branch and full commit SHA that the caller already verified against its ticket.

Name. A Planning Baseline branch is the exact <name>; never derive a replacement. Otherwise derive <name> from what $ARGUMENTS describes — treat it as a description of the work, not as a literal name, unless it already reads like a slug. Without one, derive it from the task this session is about to start: what the user just asked for, the plan under discussion, or the uncommitted changes in Context. Keep <name> short, ASCII, and valid as a branch name (git check-ref-format --branch <name>), e.g. feat/create-worktree-command, fix/null-deref. If neither the arguments nor the session says what the work is, ask the user one short question.

Where. Always ~/code/worktrees/<repo>/<name>, where <repo> is the repository name — the directory name of its main working tree, not of the worktree you may currently be in. A / in <name> just nests another level. Never place a worktree inside the repository's own working tree.

Select the source. Fetch origin, then choose exactly one path:

1. Matching registered worktree — when one registered worktree already checks out <name>, verify its branch and, when supplied, that the Planning Baseline SHA is an ancestor of its HEAD. Reuse that path. A different registered branch at the target path is a collision.
2. Planning Baseline branch — require the full SHA to exist and be an ancestor of the selected branch HEAD. If the local branch exists and is free, attach it with git worktree add <path> <name>. If only origin/<name> exists, create it with git worktree add --track -b <name> <path> origin/<name>. When both refs exist, fast-forward a free local branch with git branch -f <name> origin/<name> only when it is strictly behind the remote; keep unpushed local commits when the remote is its ancestor; divergent refs stop the operation.
3. New branch — default to origin/<default-branch>. Use current HEAD when the task depends on commits that exist only there. Use another existing base when the task requirements provide a concrete reason; verify that base resolves to a commit after the fetch, then report the selected base and reason. Uncommitted changes do not cross this boundary; report the dependency before creating anything.

Before creating. Read Context first:

• Path collision — if a non-empty target directory is not the matching registered worktree, stop. Do not choose a variant path.
• Branch already checked out — a branch can live in only one worktree at a time. If <name> is checked out elsewhere, say where; don't force it.
• Baseline conflict — a missing SHA, failed ancestry check, or divergent local/remote branch stops the operation. Do not choose another base or branch.
• Base conflict — conflicting requirements or several equally plausible bases stop the operation. Do not pick one arbitrarily.

Create a new branch and worktree in one step with git worktree add -b <name> <path> <base> so they cannot drift apart. Attach or track existing branches with the exact commands in the source-selection rules above.

Once it exists, switch into it if the session can, and report the path, branch, source (<base> and reason, Planning Baseline <full SHA>, or matching registered worktree), and whether the worktree was created or reused. Then pick up the task that prompted it.

$ARGUMENTS