What problem does it solve? When two parallel sessions open PRs targeting the same issue, handoff prompts often recommend a winner using merge-mechanics rules (first-mover, clean-against-main, single-reviewer APPROVE) that ignore implementation quality, silently shipping the inferior fix and losing substantive improvements. ## Core Features & Use Cases - Pre-Tiebreaker Delta Audit: Runs gh pr diff on both PRs and classifies each delta as substantive (new error paths, validation, different semantics) or cosmetic (naming, whitespace, comments) before accepting any tiebreaker default. - Stale-State Refresh: Re-polls gh pr view --json state,mergeable,statusCheckRollup to detect already-merged PRs via post-merge workflow signals before deciding. - Override & Documentation Workflow: Surfaces override decisions to the user in auto-mode, closes the loser with neutral coordination framing, and records the deltas in MEMORY.md and tracker entries. - Deployed-Fix Variant: Handles the case where one fix is already live by establishing ground truth from the deployed artifact and reconciling via an atomic swap instead of a blind deploy. - Use Case: A handoff prompt says "merge PR #542, it's first-mover and clean"; you diff both PRs, find the later PR #553 has a real correctness fix, and override the default to ship the better implementation. ## Quick Start Ask the AI to compare the diffs of the two parallel PRs for the same issue and verify which implementation is substantively better before merging the one the handoff prompt recommended.
