disk-cleanuplisted

# disk-cleanup — Weekly Recovery A multi-worktree development setup with 7+ named persona worktrees, each running its own `cargo build`, accumulates disk usage fast. This skill is the periodic recovery routine — safe by default, runs `--dry-run` unless explicitly told `--apply`, and never deletes branches with unmerged work. Cadence: weekly to biweekly, manually invoked. No scheduler. ## What it does 1. **Worktree inventory.** Lists all `worktrees/*-<persona>` directories under `<workspace>/worktrees/` and their disk usage. 2. **`cargo clean` per worktree.** The biggest space hog in a Rust-heavy workspace. Each worktree typically holds 2-8 GB of `target/` debug artifacts. 3. **Merged-branch cleanup.** Lists local + remote branches that have been merged to `origin/main`. **Never deletes** branches with unmerged commits. **Never deletes** branches with active worktrees. 4. **Node-side caches.** `npm cache clean --force` and `pnpm store prune` (only if the relevant tools are installed). 5. **Optional Docker prune.** `docker system prune --volumes` only when `--include-docker` is passed (this can wipe build caches that take a long time to rebuild — opt-in only). 6. **WSL VHD shrink prompt.** **Does not** run `Optimize-VHD` directly — WSL cannot shrink its own backing VHD while running. Instead, prints the exact PowerShell command to run as Administrator from a Windows terminal, with the VHDX path resolved. 7. **Reports freed bytes** per step a