It would be helpful if the agent had the ability to ingest and analyze images so that a user can provide a screenshot of an error and it deflect based on the image.