Indirect Prompt Injection in LangChain ReAct Agent via Malicious Markdown Documents | AI Breaking Wire