With the accelerated development of China's new power system, the widespread integration of renewable energy at high penetration levels and the significant increase in power electronic device adoption have led to power grid dynamic characteristics that exhibit strongly time-varying, nonlinear, and tightly coupled multi-time-scale features. Against this backdrop, rotor angle stability and voltage stability issues are increasingly intertwined and mutually influential, making transient stability conditions more complex. Traditional methods relying on mechanistic analysis and time-domain simulation face significant challenges in balancing computational efficiency and model complexity, struggling to meet the rapid response requirements of online analysis and real-time control. To address these challenges, a new framework for transient stability assessment and control characterized as mechanism-driven and data-enhanced is proposed. This framework deeply integrates power grid dynamic response mechanisms with advanced deep learning technologies, establishing an efficient and accurate stability evaluation model and a control strategy generation mechanism. It aims to enhance the capability for transient stability analysis and decision-making in the new power system, providing crucial technical support for secure and stable system operation.