Indirect Prompt Injection in LangChain Agent Enables Exfiltration of Sensitive Data | AI Breaking Wire