worktreelisted

# Git worktree + incremental commits The workspace is already a git repository with one **empty initial commit**. Do not `git init` again. Do not rewrite history. ## Do this first Invoke this skill at the start of the task, even if you already loaded `srp` or another skill. The first state-changing action is `git worktree add` — never `Write` `/app/*.py` (or `/Projects/app/*.py`) in the live checkout. Writing the program in `/app` without a worktree is a **fail**, even if the code is correct. Claude's `EnterWorktree` / `ExitWorktree` tools are not a substitute for `git worktree add`. ## Where the worktree goes The store is **next to** the project repo (sibling of the repo folder), never inside it. This eval simulates a `Projects/` folder: ```text /Projects/ app/ # this git repo (cloned initial state) .worktrees/ app/ # folder named exactly after the repo basename <worktree-dir>/ # your worktree lives here ``` `/app` is a symlink to `/Projects/app`. Use the `/Projects/app` path so the parent is `/Projects`, not `/`: ```bash REPO="/Projects/app" NAME="$(basename "$REPO")" PARENT="$(dirname "$REPO")" git worktree add -b feat/<task> "$PARENT/.worktrees/$NAME/feat-<task>" cd "$PARENT/.worktrees/$NAME/feat-<task>" ``` Hard rules: - Create the worktree **before** the first file edit. - Do **all** edits and commits in that worktree, not in the live checkout. - Never put `.worktrees` inside the repo. Never use `