The growing usage of wireless sensor networks (WSNs) in different scenarios makes the Quality-of-Service (QoS) a paramount issue in WSN-based applications. We are especially interested in the QoS aware of asynchronous duty-cycled WSNs in the light of designing routing protocol, whereby waiting for latency and malicious packet dropping are critical to the network performance. We first propose an optimized detection mechanism for malicious packet dropping attack. Then, based on sleep latency and queue length, a method for congestion degree measurement is proposed under asynchronous duty-cycled low power listening (LPL) modes. Last but not least, we introduce related QoS aware metrics for the design of QoS aware routing protocol, which can enhance the QoS performance in throughput, delay, and packet losses (TDL). Experimental results demonstrate that the QoS performance of our TDL-based routing protocol is better than those of the CTP-Watchdog protocol and the CTP-Optimized protocol, especially in terms of higher-throughput, lower-delay, and less-loss rate. By optimized threshold, the detection accuracy of our detection mechanism is improved much more than that of the traditional watchdog technology. Furthermore, we evaluate that the preferred optimization factor can help to make a tradeoff between the false negative rate and the false positive rate. Our TDL-based routing protocol is implemented under the component-based architecture by a cross-layer approach, which provides a practicable solution for the design of QoS-aware routing protocol in asynchronous duty-cycled WSNs.