Informed decision making is a critical activity in software development, but it is poorly supported by common development environments, which focus mainly on low-level programming tasks. We posit the need for agile software assessment, which aims to support decision making by enabling rapid and effective construction of software models and custom analyses. Agile software assessment entails gathering and exploiting the broader context of software information related to the system at hand as well as the ecosystem of related projects, and beyond to include "big software data". Finally, informed decision making entails continuous assessment by monitoring the evolving system and its architecture. We identify several key research challenges in supporting agile software assessment by focusing on customization, context and continuous assessment.