Recently, mobile agents have been proposed for efficient data dissemination in sensor networks [1-6]. In the traditional client/server-based computing architecture, data at multiple sources is transferred to a destination; whereas, a task-specific executable code traverses the relevant sources to gather data in the mobile-agent based computing paradigm. As described in [1], many inherent advantages (e.g. scalability, extensibility, energy awareness, reliability) of the mobile agent architecture make it more suitable for sensor networks than the client/server architecture. In this paper, a mobile agent is exploited in three levels (e.g. node level, task level, and combined task level) to reduce the information redundancy and communication overhead.