Worktree Start

Creates a lightweight isolated workspace using git worktree. Ideal for running 4-5 features in parallel, each managed by an independent agent session, without the disk overhead of a full clone.

What it does

1. Guards .worktrees/ is in .gitignore (adds and commits if missing)
2. Creates a worktree at .worktrees/<feature-name> on branch work/<feature-name>
3. Installs dependencies (auto-detects pnpm / yarn / npm / cargo / pip)
4. Runs a baseline check to confirm the clean starting state
5. Reports the path and next steps

Usage

/worktree-start <feature-name>

How to execute

bash .claude/skills/worktree-start/scripts/worktree-start.sh "<feature-name>"

Important

• Worktrees live inside the project at .worktrees/<feature-name> — no sibling directories
• Each worktree has its own index file (no git contention between parallel agents)
• Branch naming follows the same convention as fork-project: work/<feature-name>
• The .worktrees/ dir is gitignored so worktree contents are never tracked
• Always inform the user of the full path so they can open their agent session there
• Pairs with /worktree-finish to merge back and clean up