Self-organizing task allocation is vital for collaboration among multiple human and robotic agents to satisfy changing operational objectives and workspace dynamics. Nonetheless, current HRC systems heavily rely on predefined task configurations for human and robot agents, fail to consider manufacturing requirements from diverse operation sequences and varying mechanical components. To tackle this challenge, this chapter introduces a temporal subgraph approach for the task planning in Proactive HRC settings with multiple agents. The task allocation strategy is represented using a tri-layer knowledge graph that captures the relationships among tasks, agents, and operations. Simultaneously, we incorporate a temporal subgraph reasoning mechanism to extract implicit and historical knowledge from the knowledge graph, anticipating forthcoming actions for both humans and robots. To showcase the efficacy of the proposed methodology, we applied it to an assembly tasks of car engine and gearbox, respectively. The experimental results achieved significant performance.