n8n - 💡(How to fix) Fix Security: OWASP ASI06 memory poisoning defense for n8n AI agent nodes

Memory Poisoning Defense for n8n AI Agent Nodes (OWASP ASI06)

Hi n8n team 👋

n8n's AI agent nodes are increasingly used for autonomous, multi-step workflows. As these agents read from and write to memory stores (chat history, vector stores, tool outputs), they become vulnerable to ASI06 — Memory Poisoning from the OWASP Top 10 for Agentic Applications 2025.

The Threat

An attacker can inject malicious instructions into an agent's memory that persist across workflow runs, silently altering future behavior without triggering any LLM-level safety filter.

Proposed Integration

OWASP Agent Memory Guard is an official OWASP reference implementation that provides a scan-before-write pattern for any memory store.

from agent_memory_guard import MemoryGuard

guard = MemoryGuard()
result = guard.scan(content_to_store)
if result.is_safe:
    memory_node.write(content_to_store)

Install: pip install agent-memory-guard

Ask

Would the n8n team consider:

1. A security note in the AI Agent node documentation about ASI06?
2. An optional memory validation step in the agent node configuration?

Happy to draft a docs PR or prototype a custom n8n node wrapping the guard.