llamaIndex - 💡(How to fix) Fix [Feature Request] Memory Poisoning Protection for LlamaIndex via OWASP Agent Memory Guard

Problem

LlamaIndex agents using memory (ChatMemoryBuffer, VectorMemory) are vulnerable to memory poisoning attacks — adversarial inputs injected into conversation history or vector stores can cause agents to leak secrets, ignore instructions, or behave maliciously. OWASP identifies this as a top risk for LLM applications.

Proposed Solution

OWASP Agent Memory Guard (AMG) is an open-source Python library that wraps any memory store as a transparent security layer:

• pip install agent-memory-guard
• Scans every memory write for prompt injection, PII leakage, and tampering
• 92.5% detection rate on AgentThreatBench benchmark
• Designed to integrate with any BaseChatMessageHistory or similar interface

Links

• GitHub: https://github.com/OWASP/www-project-agent-memory-guard
• PyPI: https://pypi.org/project/agent-memory-guard/
• Benchmark: https://pypi.org/project/agent-threat-bench/

Would the LlamaIndex team consider adding AMG as a recommended security layer for agent memory? Happy to contribute a PR with a LlamaIndex-native integration.