Manage git worktrees for isolated per-issue/per-PR work. Three sub-actions, each in its own sibling file.
Action selection
| User intent | Sibling | Slash command |
|---|---|---|
| Create one worktree for an issue or PR | create.md |
/create-worktree <num> |
| Remove a specific worktree (interactively) | remove.md |
/remove-worktree |
| Batch-clean all completed worktrees | prune.md |
/prune-worktree |
Match the user's request to one of the three rows, then read the corresponding sibling for the full procedure.
Conventions (apply to all three actions)
- Worktree path:
../<reponame>-issue-<num>-<title-slug>/— sibling to the source repo, never inside it - Branch name:
issue-<num>-<title-slug> - Title slug: lowercase, alnum + hyphens, derived from
gh issue view --json title - Identifier rule:
issue-<NUM>-prefix is the trusted issue marker on a branch —fix/2fa-logincontains a number but is NOT an issue branch - Default branch:
git remote show origin | grep 'HEAD branch'(handles repos withmainvsmastervs custom) - State file:
${CLAUDE_PLUGIN_ROOT}/scripts/worktree-state.shregisters the active worktree so the pre-tool-use hook blocks accidental edits to the source repo
Prerequisites (all three)
ghCLI installed and authenticated- Run from inside a git repository (any worktree of the repo, not just the main checkout)
Further Reading
create.md— full create procedure (PR vs issue detection, env-file copy, state registration)remove.md— interactive selection + safety-check matrix (issue closed? branch merged? uncommitted changes?)prune.md— batch evaluation, classification, confirmation, removal