除錯
Multi-Model Review
Orchestrates a multi-model review workflow using Claude as orchestrator and Codex (via MCP) as a second-opinion reviewer. Produces adversarial, constructive, or debate-style critiques with a synthesis that highlights where the models diverge — which is where the real signal is. Use this skill whenever the user says anything like "review this with Codex", "get a second opinion", "red team this", "have Codex critique this", "multi-model review", "ask Codex what it thinks", "steelman this proposal", "debate this", or "review this code with another model". Triggers any time the user wants two AI perspectives on a proposal, draft, design, or codebase.