[Bug] Overly restrictive safety filtering blocks legitimate security remediation workflows

Bug Description
I was doing a defensive security review of my own company's ASP.NET API (multi-tenant customer portal) with Claude Code. Claude found authorization gaps in our own code and I asked it to create bug tickets with Dutch descriptions in our own Azure DevOps sprint, assigned to me, so we can fix them.

The response safety check stopped this several times: when creating the bugs with descriptions, when I asked for only descriptive titles, and when I asked for the text so I could paste it in myself. Earlier, a local reproduction against our own code (fake services, no real data or accounts) was also stopped.

This is our own codebase and our own internal ticket system; the goal is remediation, not exploitation. Claude normally does this kind of work for us without problems. Please review whether this is a false positive, and how we can do security reviews of our own code without being blocked. And this used to work just fine last week, since today this doesnt work anymore.

Environment Info

• Platform: win32
• Terminal: xterm-256color
• Version: 2.1.288
• Feedback ID: b304b349-1ee6-4cbb-b7e6-042c15fc343d