Security: Memory poisoning protection for BabyAGI task memory

Feature Request: Memory Security Layer

Hi,

BabyAGI pioneered the task-driven autonomous agent pattern that inspired an entire generation of AI agent frameworks. As BabyAGI and its derivatives are used in production, the memory/task list is a critical attack surface.

The threat: A malicious task result or tool output can be written to BabyAGI's task memory and influence all subsequent task generation and execution — a persistent, cross-session attack.

OWASP Agent Memory Guard is the reference implementation for OWASP ASI06 (Memory Poisoning):

• Scans memory reads/writes for injection payloads and structural tampering
• SHA-256 integrity baselines
• Policy-driven enforcement (block / warn / strip)
• Zero external dependencies

pip install agent-memory-guard

GitHub: https://github.com/OWASP/www-project-agent-memory-guard

Would love to discuss adding a security wrapper around BabyAGI's memory operations.