What problem does it solve? Iterating between implementation and code review often turns into an unbounded loop where LLM reviewers never run out of advisory findings, causing code to be shaped to please reviewers instead of users. This Skill gates only on high-severity blocking findings and enforces a hard cap on fix rounds, so implementation-review cycles terminate with objective stop conditions. ## Core Features & Use Cases - Bounded loop orchestration: Composes the implement and local-review skills by reference, adding only loop control, a findings ledger for cross-round recurrence detection, and a final report. - Severity-based gating: Only high-severity findings block the loop (medium only on explicit user request); advisory findings pass through to the final report untouched. - Fresh reviewers and evidence: Every review round dispatches new reviewer sub-agents, and every fix round ends with a parent-run full test suite before the next review. - Use Case: Ask the agent to implement a new feature and iterate until review findings are resolved; the loop implements, reviews, fixes blocking findings across at most 2 fix rounds, then reports remaining findings and suggests committing via the git-commit skill. ## Quick Start Use the implement-review-loop skill to implement this feature and iterate until the review findings are resolved.
