Security: OWASP Agent Memory Guard for memory poisoning defense (ASI06)

[vgudur-dev]

Feature Request: OWASP Agent Memory Guard Integration

What is Agent Memory Guard?

OWASP Agent Memory Guard is the official OWASP reference implementation for ASI06 - Memory Poisoning from the OWASP Top 10 for Agentic Applications.

Why LangMem Needs This

LangMem's memory management system (semantic memory, episodic memory, procedural memory) is exactly the attack surface that memory poisoning targets. Malicious content stored via LangMem can persist across all future agent runs.

Integration Example

pip install agent-memory-guard langchain-agent-memory-guard

from langmem import create_memory_store
from langchain_agent_memory_guard import MemoryGuardMiddleware

# Wrap LangMem store with security middleware
store = create_memory_store()
secure_store = MemoryGuardMiddleware(store, policy="strict")

# All writes are automatically scanned
secure_store.put(namespace, key, {"content": user_message})

Request

Would the LangMem team consider adding a security section covering memory poisoning (OWASP ASI06) and referencing Agent Memory Guard?

• GitHub: https://github.com/OWASP/www-project-agent-memory-guard
• PyPI: https://pypi.org/project/agent-memory-guard/
• LangChain integration: https://pypi.org/project/langchain-agent-memory-guard/

Apache 2.0 licensed.