Design Patterns for Securing LLM Agents against Prompt Injections - Explained Simply | ArXiv Explained