As the intelligence in autonomous systems expand, so too have their roles. These complex systems, and the intelligent agents of which they are comprised, serve more independent and interactive roles with humans than tools of the past. In order to ensure humans can effectively interact with this technology in human agent teams (HAT) there is a need to understand how the varying levels of autonomy and types of agent embodiment influence this interaction.   The recent technological advancements in autonomous systems blur previous delineations between distinct types of agents based on their embodiment (e.g., physical—or tangible such as warehouse robots, virtual—or digital based agents such as a virtual assistant; and embedded—or invisible agents that operate without any embodiment such as a global positioning system (GPS) assistant; Glikson & Woolley, 2020) and may exhibit multiple embodiments. For example, unmanned aircraft systems (UAS), which are physical, are also shown virtually on control stations, and exhibit embodied capabilities including object detection and mission planning. This paper and presentation will focus on what is needed to effectively train humans and manage HATs, including identification of the skills that humans need to work with agents across varying levels of autonomy. The paper will map out a framework for taskwork and teamwork skills by intersecting research on human-human team composition, selection, and training with research on HATs to address which skills are needed for which types of agents. These concepts will be discussed across emerging advanced agents that break the modern mold of singular embodiment and pose unique challenges, such as UAS that are controlled from thousands of miles away, and autonomous cars that a human must trust as a passenger traveling at highway speeds. Issues with current frameworks and areas of future research will be discussed.