A simulation framework, WMC (for Work Models that Compute) has been developed to model the complex, heterogeneous dynamics of systems that include physical systems, humans, and automated agents whose behaviors depend on situation. Unlike human-computational performance models, the WMC work model describes the actions by which the agents collectively interact with the environment to achieve their task goals. The simulation engine parses the work model, passing actions to the agents when they need to be executed, rather than needing each agent model to encapsulate its own description of task. This model form is not intended to predict or describe individual elements of human cognitive behavior within isolated tasks, but instead to describe human agents managing a range of tasks when situated within operational context. The default perfect agent executes actions exactly, but this paper elaborates on those aspects of "first-principles" agent models particularly suited to this type of simulation, including the modeling of agents' ability to monitor their taskload, actively manage competing tasks to interrupt and delay on-going actions to meet the immediate demands of the work, predict and prepare for upcoming actions, and perform internal mental simulations. In addition, based on our recent studies of how human experts situated in a dynamic environment adjust their patterns of cognitive activity in response to context, this paper discusses how strategies may also be modeled to represent the organization of actions within a foreseeable event horizon. These constructs are addressed in this paper by demonstrating how they can be modeled conceptually and then specifically implemented in dynamic simulations.