How can we observe how people respond to consequential errors by an artificial agent in a realistic yet highly controllable environment? We created a threat-detection house-search task in virtual reality in which participants form a Human-Agent Team (HAT) with an autonomous drone. By simulating risk, we amplify the feeling of reliance and the importance of trust in the agent. This paradigm allows for ecologically valid research that provides more insight into crucial human-agent team dynamics such as trust and situational awareness.