The Architecture Tradeoff Analysis Method (ATAM) is an architecture evaluation technique currently evolving at the Software Engineering Institute (SEI). ATAM has been applied to a number of command and control, real-time, and information systems. As collaborative, autonomous agents become a significant software technology, the demand for evaluating the quality attributes of the architectures of agent-based systems will increase. Very broadly, agents may be thought of as software entities that have the ability to undertake action autonomously in their particular embedded environment, according to a typically general set of requests or desired goals, and that are able to communicate with other agents as determined by their own initiative. Given an agent-system architecture, we need scenarios that could be applicable for conducting ATAM evaluations on instances of that agent architecture. This report identifies a few features in agent-based systems that could be used to classify agent-system architectures and to guide the generation of scenarios applicable to these architectures.