Show HN: OWASP Agent Memory Guard – Stop AI Agent Memory Poisoning

Category: security

Tags: ai-security, memory-poisoning, python-library

Score: 6.8/10 (Innovation: 7, Technical: 6, Documentation: 7, Utility: 7)

OWASP Agent Memory Guard provides a runtime defense layer that screens AI agent memory operations (read/write) to prevent memory poisoning attacks, including prompt injection, secret leakage, and integrity tampering. Its innovative aspect lies in addressing a less-covered attack surface (agent memory) versus typical front-door input filtering, and its flexible YAML policy engine plus drop-in middleware for LangChain make it practically useful for developers building secure AI agents.

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

Repository: https://github.com/OWASP/www-project-agent-memory-guard · Python · NOASSERTION · 17 stars

View on Hacker News