[Bug] Anthropic API Error: Overly aggressive security filtering blocks legitimate defensive code review requests

Bug Description
False positive. I was running a defensive security review on my own codebase (a staff management web app plus some Roblox game scripts I develop). The goal was to find weak spots and fix them before release, not to attack anything. Claude had already mapped the project and verified findings, and it got flagged at the very end while checking line references for the report. It happened twice in a row, even after I hit continue. My prompt asked for an exploit example for each finding, which probably set it off, but it was all about my own code and how to patch it. It would really help if reviewing your own repo didn't get blocked like this. Request ID: req_011CfNZbkaVXcYJCxkj4WRHC

Environment Info

• Platform: win32
• Terminal: vscode
• Version: 2.1.281
• Feedback ID: 7a87f6b8-73e2-4b94-8086-a39610e91438

Errors

[]