What problem does it solve? Manually reviewing pull requests, applying fixes, and re-reviewing is repetitive and error-prone. This Skill runs an autonomous review-fix loop on an open PR, alternating a fresh reviewer subagent and a fixer subagent until findings reach zero, while enforcing round and cost budgets. ## Core Features & Use Cases - Autonomous review-fix loop: Alternates reviewer and pr-resolver fixer subagents with full, delta, and confirm focus progression until consecutive clean passes are met. - Budget and round caps: Tracks estimated tokens and USD per round, blocks launches projected to cross caps, and supports budget-only unlimited-round mode. - Policy-based triage: Auto-fixes unambiguous findings, defers low-severity or diminishing-returns items, and escalates only genuine design or policy ambiguity. - Durable closed-findings ledger: Persists fixed, accepted, and deferred findings per PR so later runs do not rediscover or reverse prior fixes. - Use Case: Open a PR branch, invoke the loop, and let it review, fix lint and logic defects, re-verify the fix surface, and produce a summary canvas with findings, commits, and cost totals. ## Quick Start Run the PR review loop on the current branch's open pull request with a budget of $1.50 and a maximum of 3 rounds.
