Skill UI

Adversarial double-blind audit of a paper's .bib entries: one agent extracts each reference verbatim, a second independently re-fetches it from the web, and a third emits a field-by-field diff to catch dropped authors, wrong years, venues, and bad DOIs/arXiv ids.

This skill ensures AI models generate complete, unabridged code and files without placeholders, TODOs, or skipped sections. It enforces strict completeness for deliverables like full implementations or exhaustive lists. It prevents shortcuts like '...' or descriptions instead of code, ensuring every requested item is fully realized within token limits.