What problem does it solve? When a coding agent crashes, times out, or overlaps with another agent on the same Git checkout, you are left with a dirty tree, unclear ownership, and no trustworthy record of what actually happened. This Skill provides a disciplined recovery workflow that preserves valid work, restores a single trustworthy writer, and verifies completion instead of trusting exit codes. ## Core Features & Use Cases - Concurrent-writer arbitration: Enumerate live processes by full command, classify writers versus read-only reviewers, stop duplicate writers, and enforce one checkout, one writer. - Dirty-tree preservation and handoff: Inventory staged, unstaged, and untracked changes, preserve exact diffs and hashes, and create bounded handoffs with explicit file allowlists for a fresh worker. - macOS cloud-sync recovery: Detect dataless File Provider placeholders that corrupt checkouts and relocate work to a stable non-synced checkout. - Verified closeout: Re-run controller gates after final commits, bind desktop package identity to source, and reject self-reported PASS without independent evidence. - Use Case: An agent exits mid-task leaving uncommitted changes and a possibly orphaned background worker. Use this Skill to freeze new writers, inventory the checkout, preserve the diff, hand off to a fresh model with exact remaining blockers, and verify the final commit. ## Quick Start Ask the agent to recover the interrupted coding work in this Git checkout, verify no other writer is active, and preserve the dirty tree before continuing.
