← ClaudeAtlas

quorumlisted

Convene three independent reviewers from different models and reconcile their findings into one verdict with scores per round. Use when the user says "quorum", asks for a panel, a council, a second and third opinion, or a multi-model review.
Udhay-Adithya/skills · ★ 0 · AI & Automation · score 70
Install: claude install-skill Udhay-Adithya/skills
# Quorum Three reviewers, each a different model, each at high reasoning effort. Their disagreements are the point: one model's blind spot is rarely shared by all three. ## Before convening Address outstanding PR comments, fix known issues, and run the relevant checks first. The panel reviews the updated implementation; do not defer known fixes until reviews come back and waste a round on findings already known. The exception is when the implementation is about to be discarded and the question is which direction to take instead. Then skip fixes to code being replaced. ## Picking the panel Use one strong model from each of three different providers. If the request names members, use those. Otherwise take the current flagship from each provider. In Claude Code, start one subagent per model with the `Agent` tool, in parallel, setting `model` per agent and `effort` to high. Where a provider is not reachable from this harness, say so rather than substituting a second model from a provider already on the panel. A panel of three where two are the same model is not a panel of three. Give each reviewer the request and the context it needs. Keep their work read-only for a review or a critique. ## Running a round Ask each reviewer for findings and a score out of 10. One full review pass is one round. A fresh review after fixes starts a new round. Collect the responses and reconcile them. Report the useful findings, where the reviewers disagreed, and the recommended next step.