SystemC has emerged as a de facto standard modeling language for hardware and embedded systems. However, the current standard does not provide support for temporal specifications. Specifically, SystemC lacks a mechanism for sampling the state of the model at different types of temporal resolutions, for observing the internal state of modules, and for integrating monitors efficiently into the model's execution. This work presents a novel framework for specifying and efficiently monitoring temporal assertions of SystemC models that removes these restrictions. This work introduces new specification language primitives that (1) expose the inner state of the SystemC kernel in a principled way, (2) allow for very fine control over the temporal resolution, and (3) allow sampling at arbitrary locations in the user code. An efficient modular monitoring framework presented here allows the integration of monitors into the execution of the model, while at the same time incurring low overhead and allowing for easy adoption. Instrumentation of the user code is automated using Aspect-Oriented Programming techniques, thereby allowing the integration of user-code-level sample points into the monitoring framework. While most related approaches optimize the size of the monitors, this work focuses on minimizing the runtime overhead of the monitors. Different encoding configurations are identified and evaluated empirically using monitors synthesized from a large benchmark of random and pattern temporal specifications. The framework and approaches described in this dissertation allow the adoption of assertion-based verification for SystemC models written using various levels of abstraction, from system level to register-transfer level. An advantage of this work is that many existing specification languages call be adopted to use the specification primitives described here, and the framework can easily be integrated into existing implementations of SystemC.