Minimal blueprint to probe whether model outputs reflect specified real-world prevalence for underspecified scenarios. Uses simple weighted matches (no JS) on a structured tag line appended to each story.