What problem does it solve? High-stakes changes like plans, implementation diffs, and release closeouts need independent review evidence, but relying on a single model or vendor creates blind spots. This Skill runs a customizable cross-vendor advisor board through the agent-harness runtime so every critical artifact gets structured multi-vendor scrutiny. ## Core Features & Use Cases - Availability-aware board composition: Seats a 4-reviewer board (Claude Fable 5, Grok 4.6, GPT-6 Astra, Gemini 3.8 Flash) with distinct review lenses, backfilling seats onto available vendors so the board never collapses below 3 reviewers. - Three material-feeding modes: Pass artifacts inline, stage local files via artifact_ref/brief_ref, or use true by-reference context_refs for large or private files that legs open with their own tools. - Heartbeat-based liveness and bounded loops: Reclaims stalled legs via heartbeat extinction rather than wall-clock timers, supports optional per-leg hard deadlines, and structures review-fix-re-review loops with delta reviews and round caps. - Use Case: Before merging a risky refactor, run phase-loop advisor-board diff.md to collect AGREE/DISAGREE verdicts from four independent vendor seats, then re-run only dissenting seats against the delta after each fix round. ## Quick Start Run the advisor board on my change by executing phase-loop advisor-board with the path to my review bundle and report each seat's verdict.