Knowledge graphs have become an indispensable technology for organizing and processing vast amounts of information, with applications spanning intelligent robotics, risk management, recommender systems, and healthcare analytics. However, the effectiveness of knowledge graphs in downstream tasks is often limited by inherent structural incompleteness. To address this issue, we propose TS-TKGC (Two-Stage Temporal Knowledge Graph Completion), a novel reinforcement learning-based method. Our TS-TKGC method consists of two key stages: clue search and temporal reasoning. In the first stage, reinforcement learning is employed to identify informative clues. In the second stage, the method integrates Gated Recurrent Units (GRU) for temporal reasoning, alongside a multi-dimensional reward mechanism to optimize the training strategy. Finally, experimental results validate the feasibility and effectiveness of the proposed key technique, demonstrating the model’s capability to enhance temporal knowledge graph completion.