In Socially Assistive Robot (SAR) development, both functionality and user perception of the robot’s actions and nature are crucial. While existing literature identifies desired perception goals for SARs, translating these abstract concepts into concrete technical requirements remains challenging. We address this gap through our experience developing SemuBot, an open-source SAR for social skills training. Our approach integrates domain goals, user functional and emotional objectives, robot perception requirements, and technical specifications into a structured requirements framework. Through iterative refinement, we developed four different requirement representations, with technical teams favoring categorization by robot functional requirements. This methodology provides a practical bridge between high-level perception goals and implementable technical specifications, particularly valuable for open-source SAR development.