The ubiquitous presence of autonomous systems has given rise to the need for verification to ensure transparency and trustworthiness of such systems. Traditional methods that treat autonomous systems as monolithic are not tractable given the complexity of in situ vehicle behavior in a stochastic environment. To assist with the problem of verification, we propose a new ontology, the form of a hierarchical abstract modeling language for autonomous system capabilities. This abstraction is useful in that it establishes a conceptual model describing capabilities of autonomous systems and the relationships between them to represent complete systems. Building upon this conceptual model, we develop a mathematical framework that describes autonomous systems as graphs of capabilities capable of analysis and develop a metric for the space of verified capabilities. These mathematical tools enable a formal, analytical comparison of multiple systems. With the mathematical framework in place, we demonstrate its use by considering and simulating a simple example involving an autonomous vacuuming robot.