What problem does it solve?
Starting a task in an isolated git worktree normally requires manual branch naming, worktree creation, and directory switching, and doing it from an AI session involves harness-specific confirmation prompts and failure modes. This Skill automates the full create-and-enter flow for worktrunk worktrees, including cross-repo targets and recovery from common errors.
Core Features & Use Cases
- One-call worktree creation and entry: Creates a new worktree via the
wt CLI and re-roots the session into it, using EnterWorktree by name when possible to skip confirmation prompts.
- Cross-repo support: Accepts an optional repo path argument to create the worktree in a different repository, with reachability checks and escalation guidance when the target is outside allowed directories.
- Error-driven recovery: Handles existing-branch conflicts, already-in-worktree sessions, and denied confirmations with concrete fallback steps, and carries uncommitted work across via git stash.
- Use Case: A developer asks the assistant to fix a parser bug in its own worktree; the Skill picks a branch name, runs
wt switch --create, enters the worktree, and performs the task there, leaving cleanup to wt merge or wt remove.
Quick Start
Ask the assistant to create a new worktree for a branch and switch into it, for example by invoking /wt-switch-create my-feature -- fix the parser bug.
