Systems operating in hazardous and remote environments is both desirable from a scientific perspective and incredibly difficult from an engineering and logistical perspective. This paper develops a novel two-system top-down computational-cognition-based agent and investigates its ability to respond to unanticipated situations in a hazardous environment. A simulated space mission is performed in which a Martian settlement site is constructed by a team of robots. In order to evaluate the agent’s ability to respond to unanticipated mission scenarios, “black swan” events are added to the simulation. We found that the agent was able to respond well to black swan events that changed the mission conditions in a detectable way, but black swans that were either undetectable or gave the agent inaccurate mission information increased the likelihood of failure. This work presents an interesting first step towards autonomous systems that are more resilient when facing hazardous or unknown environments. Additionally, the techniques presented here can handle large, complex problems with very minimal computational resources.