What problem does it solve? AI-generated text can look like sound reasoning while containing invented facts, fake citations, and silent omissions, and asking the model to explain itself only produces more confident text. This Skill tests the output itself instead of trying to walk an author back through reasoning that never happened. ## Core Features & Use Cases - Eleven Targeted Probes: Detect hallucinated citations, framing bias, false balance, vague sections, and silent omissions, selected based on what being wrong would cost. - Fresh Conversation Strategy: Recommends restarting with a short brief instead of correcting a contaminated thread, avoiding anchored errors. - External Verification Discipline: Directs every check to primary documents, real data, or people with direct knowledge rather than trusting a second model's agreement. - Use Case: You pasted a ChatGPT answer containing a statistic and a citation into a report. The Skill asks one probe at a time, identifies the claims that would collapse the answer if false, and tells you exactly where to verify each one. ## Quick Start Paste the AI-generated answer you are about to rely on and ask the assistant to verify this AI output before you use it.
