Unmanned Surface Vehicle swarm combat will be the main tactical mode in the future. Based on the requirements of multi-USV executing combat missions, agent-based USV swarm combat overall framework is designed and analyzed in details. Then we investigate USV offensive and defensive strategies for executing different mission tasks. With the designed USV swarm decision-making rules, each individual USV can perform strategies such as following, avoiding, striking and cooperating. The feasibility and effectiveness of USV swarm combat decision making methods was verified through simulation experiments.