We want to be able to instruct the AI Agent to respond differently to different message types (like "text", "image", "video", "audio", "file", "share", "story_mention", "ig_reel")