This paper presents VILLAIN, developed by Team HUMANE, a multimodal fact-checking pipeline that verifies claims through multiagent collaboration. The system employs multiple vision–language model agents across all stages of fact-checking: evidence analysis over different modalities, question–answer generation, and verdict prediction. As a result, VILLAIN ranked first in the shared task of the Ninth FEVER Workshop, achieving a veracity score of 0.546 on the test set.