What problem does it solve? Unverified claims about root causes, completed work, or irreversible actions often reach decision-makers without independent scrutiny, leading to repeated failures and broken trust. This Skill enforces an adversarial verification protocol that treats every consequential claim as false until proven otherwise. ## Core Features & Use Cases - Refute-by-default verification: Challenges claims using fresh-context intake (claim + evidence only, never the author's reasoning) to avoid inheriting the author's blind spots. - Source-backed refutation: Searches Context7, Exa, fuse-browser, and actual code via Read/Grep/Glob to find concrete failure scenarios instead of relying on memory. - Bounded protocol with mandatory verdicts: Runs a maximum of 2 refutation rounds, then always outputs a structured verdict (CONFIRMED / REFUTED / UNCERTAIN) with sources checked and recommendations. - Use Case: Before telling a project owner "the bug is fixed and deployed," run this protocol to hunt for a counter-example in the actual code and documentation, then attach the verdict to the report. ## Quick Start Challenge the claim that the login fix resolves the timeout issue by verifying it against the actual code and documentation before reporting it as done.
