The internet is flooded with examples of the impressive capabilities of chatbots like ChatGPT, but it can be difficult to know exactly what they can and can't do as it changes every year (or every month!). In this workshop, we will briefly present some of the tests that computer scientists use to track AI capabilities. Participants will work with each other in pairs to limit-test chatbots or other AI systems with "prompt engineering." One person will come up with a challenging task, and the other will be a "prompt engineer" who tries to accomplish that task with a clever use of prompting. We will conclude with a brainstorming session for new chatbot applications based on what we learn.