← ClaudeAtlas

cleanuplisted

Use when the user wants to clean up, tidy, or reset a GitHub repo — even if they name only one part. Triggers: repo cleanup, repo housekeeping, prune merged branches, delete stale remote branches, tidy my worktrees, drive all open PRs to merge-ready, wipe build and cache artifacts, reset local checkout to match remote.
Tamircohen28/tamirs-superpowers · ★ 2 · AI & Automation · score 74
Install: claude install-skill Tamircohen28/tamirs-superpowers
## Live context !`git rev-parse --show-toplevel 2>/dev/null && echo "repo root: $(git rev-parse --show-toplevel)" && git remote get-url origin 2>/dev/null || echo "not a git repo"` !`gh auth status 2>&1 | head -1 || echo "gh: not available"` # repo-cleanup Full repository housekeeping: prune remote branches, drive open PRs in parallel via sub-agents, clean local worktrees, discard or rescue uncommitted work, and reset the local environment to a clean state matching remote. ## Why this skill exists Repos accumulate cruft quickly — merged branches that weren't deleted, PRs that stalled, worktrees from old tasks, temp build files, and local checkouts that drifted from `main`. Doing this manually is error-prone and tedious. This skill sweeps all of it in one pass, asks for confirmation before any destructive action, and parallelizes PR work across sub-agents so the whole process finishes faster. **Why this skill is model-invocable** (`disable-model-invocation: false`): cleanup is meant to be *orchestrated* — an operator points sub-agents (or a Workflow) at many repos and fans out one cleanup per repo. Gating a skill with `disable-model-invocation: true` also blocks that path, because a sub-agent invoking a skill is itself model invocation. So safety lives **inside** the skill — every destructive step shows a plan and waits for confirmation, and dry-run mode changes nothing — not in the invocation flag. The flag is reserved for skills that must *never* be invoked autonomously