Indirect Prompt Injection in AI Agents via Unsanitized Web Content Leading to Data Exfiltration | AI Breaking Wire