Device agents are typically used through IM tools and rely on remote desktop control, system-level API calls, or screen recognition technology; they have full permissions over the entire device, with a low barrier to entry but high risk.