Show HN: Agent Memory Guard – OWASP defense for AI agent memory poisoning

Category: security

Tags: ai-security, memory-poisoning, owasp

Score: 7.5/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 8)

Agent Memory Guard is an OWASP reference implementation that provides a runtime defense layer against memory poisoning attacks on AI agents, screening reads and writes to prevent prompt injection, secret leakage, and integrity tampering. Its innovative combination of threat detection, policy enforcement, and forensic rollback capabilities addresses a known security gap in agentic applications with strong benchmark results.

Target audience: backend devs, ai-engineers, security-engineers

Repository: https://github.com/OWASP/www-project-agent-memory-guard · Python · NOASSERTION · 16 stars

View on Hacker News