What problem does it solve? Forward-only drafting and polishing pipelines never argue against a paper the way a real reviewer will, so weak claims, unsupported numbers, and internal inconsistencies survive into the final PDF. This Skill hardens a drafted paper by simulating an adversarial peer-review panel before figures and LaTeX finalization. ## Core Features & Use Cases - Isolated multi-lens review: N reviewers (Theory/Foundations, Empirical/Benchmark, Applied/Systems) each read the whole paper independently, and every issue must carry an exact verbatim evidence quote plus a close criterion. - Adversarial verification: Each candidate issue faces three perspective-diverse skeptics (misreading, already-addressed, scope-or-severity) and survives only if a majority fails to refute it, filtering plausible-but-wrong critiques. - Loop-until-dry with tiered execution: Review rounds repeat until no new issues survive, running identically via a Workflow script, parallel subagents, or in-context sequential role-play depending on available tools. - Use Case: After refining a research proposal draft, run this stage to get a prioritized, severity-sorted issue list (blocker/major/minor/nit), fix each item through the refine stage against its close criterion, then re-run lint checks to confirm the fixes landed cleanly. ## Quick Start Review my drafted paper in the current working directory with an adversarial reviewer panel and list the verified issues by severity.
