Sedentary behavior is a global threat to health. Increasing physical activity (PA), e.g., by walking more steps per day, can interrupt physical inactivity and is associated with a lot of positive health effects. However, individuals often struggle to integrate more steps into their daily lives. Therefore, motivating goals could serve as a remedy. They need to be challenging to nudge the individual towards more steps, yet at the same time, they should be attainable to prevent frustration. Up to now, commercial and research-based fitness devices and health recommender systems predominantly use static goals or provide rather simplistic adaptive approaches. Against this gap, we provide requirements and derive design options for adaptive step goal features in health recommender systems. We demonstrate the application of our requirements and design options to a concrete smartwatch-based recommender system that we are currently building, and finally provide a summary of our insights in the form of a checklist for designing adaptive step goal features.