We propose a two-tier hierarchical cyber-physical framework for analyzing transient stability in the smart grid in the face of cyber-physical attack. We model the smart grid as a networked multi-agent dynamical system in which each agent includes multiple generators, measurement devices and local control. The agents interact with each other through cyber-physical integrated coordination and the generators within the same agent interact through the physical couplings. We present a cyber-physical flocking-based protocol to address the issue of system faults. We extend this work to derive a spectral matrix based generator coherency identification algorithm to analyze the coherency amongst synchronized generators after fault clearing. Generators with high coherency are grouped in an agent while those with low coherency form another. For each agent, the generator having highest inertia is selected as the lead component. In order to reduce the required communication and energy of the control protocol, cyber interaction between agents is realized by coupling only lead generators. By using our developed flocking-based protocol, all the lead generators achieve consensus to the desired nominal frequency and their phase angles achieve phase cohesiveness. Subsequently secondary generators within each agent are able track these stabilizing lead dynamics due to strong physical couplings. Simulation results are presented to support the proposed framework.