What problem does it solve? Reviewing code changes and planning artifacts often stalls in open-ended feedback cycles with no clear stopping point. This Skill runs a bounded review loop that confirms or rejects findings, applies fixes, and terminates at a defined convergence state or iteration cap. ## Core Features & Use Cases - Shared review loop protocol: Drives code, plan, brainstorm, and PRD review iterations for both in-session (ptp) and external Codex reviewers, with a configurable iteration cap (default 5). - Severity thresholding: Resolves a review.minSeverity floor (low/medium/high/critical) from layered config so low-priority findings are reported but do not block convergence. - Durable convergence markers: Writes per-kind stage records (stages/code.json, plan.json, etc.) with reviewer sets, tallies, and content fingerprints so later runs can skip redundant reviews. - Use Case: After implementing a change, run /ptp:review-loop to have findings reviewed, confirmed, fixed, and tallied until the diff converges or the iteration cap is reached, with a fingerprinted marker proving what was reviewed. ## Quick Start Ask the AI to run the ptp review loop on the current OpenSpec change so findings are reviewed, fixed, and converged within the iteration cap.
