What problem does it solve? Superficial reviews miss hidden defects, gaps, and unstated assumptions in code changes, specs, and documents. This Skill applies a deliberately cynical, skeptical review pass that assumes problems exist and forces at least ten concrete findings to surface. ## Core Features & Use Cases - Adversarial Analysis: Reviews any artifact (diff, branch, uncommitted changes, spec, story, or document) with extreme skepticism, looking for what is missing as well as what is wrong. - Structured Findings Report: Outputs findings as a Markdown list of issue descriptions, requiring a minimum of ten issues per review. - Halt Safeguards: Stops and asks for clarification when content is empty or unreadable, and treats zero findings as suspicious rather than success. - Use Case: Before merging a pull request, run an adversarial review on the diff to catch missing error handling, unstated assumptions, and edge cases that a friendly review would overlook. ## Quick Start Ask the AI to perform an adversarial review of the current diff or a pasted document and list at least ten findings.