Pretty Sweet vuln we found in OneDev @ Vulnetic! A year ago models struggled to do more than read response codes, its really cool to see how far the technology has come to where our AI Sable can reason past a 403.