Show HN: We replayed 12 real AI agent incidents against our open-source tool

Category: security

Tags: ai-agents, security, guardrails, benchmark, open-source

Score: 6.0/10 (Innovation: 6, Technical: 6, Documentation: 7, Utility: 6)

Squidbrake is an open-source guardrail tool for AI agents that intercepts harmful actions by evaluating commands and sessions against security rules. It demonstrates effectiveness by replaying 12 real incidents and 62 whole sessions, showing high harmful-action stop rates with low false positives and fast decision times.

Target audience: AI engineers, security engineers, devops, and platform teams deploying autonomous agents

Repository: https://squidbrake.com/benchmark

View on Hacker News