Plant antimicrobial peptides (AMPs) are small peptides with low toxicity, essential roles in plant growth, and the ability to combat drug resistance. They show significant promise in agriculture and medicine. However, current AMP prediction methods often face shortcomings such as low efficiency, strong data dependence, and high computational costs. To address these issues, we propose PCLPAMP, a model incorporating prompt tuning and contrastive learning for plant AMP prediction. Continuous prompts guide the protein language model (LM) ProtT5 in extracting task-specific feature embeddings, while contrastive learning optimizes feature distribution to improve discrimination. PCLPAMP achieves an accuracy of 94.15%, outperforming the state-of-the-art (SOTA) model LMPred-T5XL UNI by 2.63%. Additionally, it outperforms the low-rank adaptation (LoRA) fine-tuning approach in both accuracy and computational efficiency. The complete code and datasets can be accessed at: https://github.com/Oliver4587/PCLPAMP .