What problem does it solve? Agent-generated claims, references, and deliverables often contain unverified assertions, broken URLs, or exaggerated metrics. This Skill standardizes how reviewers challenge claims, test counter-hypotheses, and report confidence levels so quality assurance output is consistent and evidence-based. ## Core Features & Use Cases - Counter-Hypothesis Methodology: For every claim, generate alternative explanations and test them against available data before accepting it. - Structured Review Template: Produces a standardized fact-check report with a claims table, evidence notes, confidence flags (Verified, Unverified, Contradicted), and a final verdict of PASS, PASS WITH NOTES, or NEEDS REVISION. - Reference Verification: Checks that URLs, package names, API endpoints, and external references actually exist. - Use Case: Before approving an architecture decision, a coordinator spawns a challenger agent to fact-check a claim like "this change saves 75% latency", requiring cited evidence within three investigation cycles. ## Quick Start Ask the agent to fact-check a specific claim or deliverable and require cited evidence for every verdict with a maximum of three investigation cycles.
