crewai - 💡(How to fix) Fix [FEATURE] Security: OWASP ASI06 memory poisoning defense for CrewAI agent memory

Feature Area

Core functionality

Is your feature request related to a an existing bug? Please link it here.

NA - This is a security enhancement request, not related to an existing bug. OWASP ASI06 (Memory Poisoning) is a newly classified threat in the OWASP Top 10 for Agentic Applications that affects all AI agents with persistent memory.

Describe the solution you'd like

Integrate OWASP Agent Memory Guard as a security layer for CrewAI's memory system.

The Problem: CrewAI agents using long-term memory, short-term memory, or entity memory are vulnerable to memory poisoning (OWASP ASI06). An attacker can inject malicious content that persists across sessions and silently alters agent behavior.

The Solution: Agent Memory Guard provides:

• Cryptographic integrity verification for stored memories
• • Semantic anomaly detection to flag suspicious entries
• • • Pattern-based heuristics for known attack patterns Integration Example:

from agent_memory_guard import MemoryGuard

guard = MemoryGuard(policy="strict")

def safe_memory_write(content):
    result = guard.validate_memory(text=content)
    if not result.is_safe:
        audit_log.record(content, result.threat_type)
        return
    memory_store.write(content)

Performance: 100% detection on direct injection, 94% on encoded payloads, <3ms latency overhead.

References:

• OWASP Top 10 for Agentic Applications
• • PyPI: agent-memory-guard
• • • GitHub Action for CI/CD

Describe alternatives you've considered

No response

Additional context

No response

Willingness to Contribute

Yes, I'd be happy to submit a pull request