A low-power wireless sensor network (WSN) implements dynamic communication protocols and embedded sensing applications on resource constrained platform. WSNs utilize dozens of tasks, which have differentiated realtime requirements. This requires an efficient implementation with the use of a real-time operating system optimized for WSNs. Current WSN operating systems are based either on preemptive or event-driven kernels. Preemption provides accurate timings but requires large data memory footprint. Event-driven have small footprint but do not support time as accurately. This paper presents a new HybridKernel for WSNs which combines the advantages of both kernels. It meets five key requirements without any major drawbacks: it halves footprint of preemptive kernels, it provides 2 us timing accuracy, it minimizes energy consumption, and it can be easily configured and used between preemptive and event-driven parts through a coherent system call interface.