The Breitbart Omar Test

At first, the agent was missing some loud rhetorical tells in this text. In particular, it often let the opening superlative slide (“largest welfare scam in history”) possibly because of the concurrent presence of "allegedly", it didn’t reliably tag in-body suspense ramps like “But that’s not all…” or “Here’s the thing…,” and sometimes treated an in-body rhetorical question (“You be the judge…”) as if it were a headline pattern.

Added a small, taxonomy-first prompt addendum that reinforces “annotate cues, not intent,” highlights superlatives as a loaded-language risk unless clearly attributed, and clarifies that in-body curiosity-gap and insinuation bridges should route to agitation/framing patterns. In the last run, captured below, those previously missed aspects were caught — especially the scale-claim cue in Segment 1 and the suspense bridge in Segment 10.