Systems Autonomous Cyber-Physical Systems (CPS), including small Uncrewed Aerial Systems (sUAS), traditionally operate within well-defined safety envelopes, continuously monitoring runtime behavior and triggering interventions when predefined constraints are approached or violated. However, as CPS evolve to incorporate higher levels of cognition and reasoning, they are increasingly required to make complex runtime decisions that, while not necessarily posing direct safety risks, can significantly impact mission outcomes. Ensuring the integrity of these decisions in dynamic and uncertain environments is critical to advancing autonomy in a way that is both effective and trustworthy. This keynote explores the challenges of achieving mission-level decision integrity in autonomous systems and introduces the concept of decision guardrails—a framework for validating and constraining autonomous decisions in real time. Operating at a higher level of abstraction than traditional safety envelopes, decision guardrails assess assumptions, strategies, and plans both before and during execution to ensure alignment with mission objectives, human expectations, and ethical considerations. Examples drawn from DroneResponse, a multi-vehicle system of autonomous small Uncrewed Aerial Systems designed for dynamic human-partnering, demonstrate how decision guardrails can enhance mission effectiveness while maintaining safety and accountability. By reinforcing decision integrity in complex and uncertain environments, these guardrails lay the foundation for the next generation of trustworthy autonomous drone swarms.