Driven by rapid urbanization and climate change, urban safety is increasingly confronted with systemic risks. These cross-domain, cross-boundary, and cross-system risks in complex urban systems fundamentally challenge traditional risk prevention and control paradigms based on fragmented management approaches, thereby necessitating the development of adaptive systematic prevention and control methodologies and technological systems. This study first identifies major challenges, including the varied spatiotemporal distribution of urban disaster risks, insufficient understanding of system complexity, inadequate linkage between early warning and emergency response, and the superficial application of digital and intelligent technologies. Based on this, it proposes a "risk perception‒disaster evolution‒early warning and response" entire-chain prevention and control framework, together with development objectives in terms of mechanism understanding, forecasting technologies, early warning capabilities, and integrated prevention and control systems. Furthermore, several key scientific issues are identified, including the coupled risk evolution mechanisms among extreme hazards, infrastructure systems, and social systems; methods for dynamic risk prediction, assessment, and critical threshold identification under extreme events; and systematic approaches for collaborative risk prevention and governance. The study further proposes key technological directions, including the coupled evolution mechanisms among extreme hazards, infrastructure systems, and social systems; an extreme disaster simulator integrating digital twin systems and social system models; theories for precise risk prevention and governance; and intelligent human‒machine collaborative platforms and applications for extreme disaster prevention and control. Using the extreme rainstorm event that occurred in Zhengzhou on July 20, 2021 as a reference scenario, we conduct a high-fidelity disaster simulation in Hefei, demonstrating the comprehensive simulation capability and scenario applicability of the proposed technical framework. Furthermore, the study looks ahead to future development directions in urban system coupled risk prevention and control, including entire-chain refined risk assessment, global systemic risk cognition, pre-event rehearsal-type decision support, intelligent collaborative risk governance, and cross-fusion research methods, thereby providing theoretical support and practical guidance for building urban safety and resilience systems.