What problem does it solve? Getting a consistent, verified code review of a change is hard when findings come from a single unchecked pass. This Skill runs one review round using Claude's own finder agents — one per angle of the chosen rung — and has every candidate finding checked by an independent verifier, returning verified findings plus a coverage record without ever modifying the code. ## Core Features & Use Cases - Angle-based finder agents: Launches one agent per review angle at the rung (medium or high) the caller fixes, over the range the caller pins. - Independent verification: Where the session can launch agents, every candidate finding within the rung's budget is checked by an independent verifier before being reported. - Review-only contract: Returns findings and a coverage record and never fixes anything, leaving decisions to the caller. - Use Case: A pipeline, batch worker, or subagent needs a review pinned to a specific commit range with verified findings handed back, or a user asks for a Claude review of the current change. ## Quick Start Ask the AI to run a Claude review of the current change at the medium rung and return the verified findings.
