[Bug] Overly aggressive cybersecurity content filter flags legitimate defensive security work

Bug Description
❯ A legitimate defensive security task got flagged as a cybersecurity topic and switched the model from Fable 5 to Opus 4.8. The work was a refactoring of an internal academy-management web app: server-side data fetching, RLS, authentication guards, rate-limiting, and PII protection — all defensive. Likely triggers were defensive-context terms such as "brute-force protection", "SQL injection vulnerability report", "open redirect prevention", "timingSafeEqual", and "attackers vs defenders" (a red-team/blue-team review framing). This is a false positive — no offensive or harmful intent. Please refine so defensive security work isn't interrupted.

Environment Info

• Platform: darwin
• Terminal: Apple_Terminal
• Version: 2.1.172
• Feedback ID: bae9627c-f0aa-4c82-838f-177eb29167f2

Errors

[]