What problem does it solve? AI coding sessions often produce plausible code, review their own work, and report success without independent verification. This Skill enforces a delivery discipline where intent is written down first, code is reviewed by six independent lenses in fresh context, and a judge adjudicates findings — with a single human approval gate as the only checkpoint. ## Core Features & Use Cases - Autonomous end-to-end delivery: Takes an approved plan and runs build, six parallel review lenses, adjudication, and a recheck of the judge's own commits without further supervision. - Single approval gate: The user approves the plan once; the pipeline records that authorization in state.json and never asks again, while detecting stale approvals from runs that died before building. - Append-only review rounds: Re-runs on a branch add new review files without deleting earlier ones, and escalations can be covered afterward by reviewing only the delta since the last verdict. - Use Case: After writing and approving a plan for a feature branch, run the pipeline once before opening a pull request so the diff is built, reviewed from six independent angles, adjudicated, and reported in verdict.md. ## Quick Start Ask the AI to run the quorum pipeline for the current branch's approved plan and report the verdict when it finishes.
