What problem does it solve? After multiple review lenses produce conflicting findings on a code change, someone must decide which findings are real, fix the valid ones, and produce an auditable verdict — this Skill performs that adjudication step instead of letting the AI that wrote the code grade its own work. ## Core Features & Use Cases - Finding adjudication: Verifies every review finding against the actual code and assigns each one a disposition of Accepted, Rejected, or Escalated with stated reasoning. - Fix application and regression gating: Applies minimal accepted fixes, runs the full test suite, and refuses to report a verdict over a red suite. - Verdict and state recording: Writes a structured verdict.md covering acceptance criteria, dispositions, escalations, and enforcement posture, then records pipeline state via guard and state scripts. - Use Case: After six review lenses flag twelve findings on a pull request, run this step to confirm three are real defects, fix them, reject the rest with reasons, escalate one API design decision to the user, and produce a verdict the reviewer can read instead of re-deriving the reasoning from the diff. ## Quick Start Run the quorum adjudication step for the work item identified by its slug to judge all review findings and produce the verdict.
