Prompt Injection Detection and Mitigation via AI Multi-Agent NLP Frameworks