What problem does it solve? Running a large batch of independent work items one at a time wastes time and stalls progress between messages. This Skill turns multi-unit batches — fleet-wide harvests, multi-repo syncs, overnight backlogs — into a persistent coordinator loop that dispatches workers, monitors progress on a heartbeat, routes feedback, and gates merges without waiting on a human trigger between steps. ## Core Features & Use Cases - Parallel worker dispatch: Spins up one background agent per unit of work, each on its own git worktree, branch, and PR, with an explicit directive to keep going until the branch is actually mergeable. - Heartbeat scheduling: Uses scheduled wakeups to re-check in-flight PRs, CI status, and stuck workers at an interval matched to the batch's pace instead of idling between notifications. - Feedback routing and merge gating: Resumes the same worker with review feedback rather than re-implementing fixes, and requires green builds, tests, and lint plus a manual diff read before merging. - Use Case: You need to apply the same migration across 15 repositories. Dispatch one worker per repo, let the heartbeat loop track each PR's CI status, route review comments back to the responsible worker, and merge each branch only after checks pass and you have read the diff. ## Quick Start Ask the AI to run this batch of independent tasks as a manager loop with one worker per repository and a heartbeat to track PR status until all are merged.
