Show HN: AgentThreatBench – Benchmark for AI Agent Memory Security
Category: security
Tags: ai-security, agent-memory, prompt-injection, owasp, python
Score: 7.8/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 8)
Agent ThreatBench is an OWASP-incubated runtime defense layer that screens AI agent memory reads and writes to block prompt injection, secret leakage, and integrity tampering. It addresses a novel security surface (memory poisoning) not covered by existing input-level defenses, with strong benchmark results and multi-framework integrations.
Target audience: backend devs, ai engineers, security engineers, devops
Repository: https://github.com/OWASP/www-project-agent-memory-guard · Python · NOASSERTION · 23 stars
View on Hacker News