<p>This article concentrates on the dynamic event-triggered-based prescribed-time consensus problem for pure-feedback multi-agent systems (MASs) in the presence of input dead-zone nonlinearities and deferred output constraints. A modified dynamic event-triggered communication approach is developed to effectively save channel bandwidth in the communication network, and the event-triggered conditions can be dynamically adjusted based on the current tracking performance of followers. In order to exclude the impact of the non-affine structure, the pure-feedback follower is restructured into strict-feedback form by employing mean value theorem. Meanwhile, some barrier Lyapunov functions (BLFs) with error transformation mechanism are formulated to achieve the deferred output constraints in the pure-feedback MASs, and a dynamic decomposition method is devised to eliminate the influence of input dead-zone nonlinearities. Finally, a two-link manipulator example is considered to demonstrate the effectiveness of proposed consensus protocol.</p>