[Feature Request] Improve Fable safety switches around authentication and authorization for complex system design

Bug Description
Your flags for cybersecurity are absolutely terrible, they flag anything remotely related to security as a downgrade to Opus. There is no way to design and build complex software and platforms without addressing authentication and authorisation. If a solution is designed without addressing security, then it is a sign it is insecure by design and should trigger a step-up, not step down.
Right now, authentication and authorisation management is often where software security fails. It's hard, complex, combines many intricate protocols and concepts than need to be wrapped to deliver a secure and manageable solution. This is where "fine-grained" is left out for convenience and manageability. This is where LLMs have potential...but have failed so far all the way up to Opus. They produce circumvoluted designs, either too complex to be managed or too weak to enforce anything, and most often reinvent the wheel.

Fable shows promise to break the vicious cycle of whack-a-mole cybersecurity has been stuck in for the last decade: spending resources on detecting and fixing defects and poor design rather than designing for safety and usability. The few cases where Claude didn't trigger a downgrade to Opus when designing my secure-by-default platform have been spot on and extremely valuable. Most models, including Opus are failing miserably at designing effective AuthZ and Policy enforcement systems, and create an absolute nightmare to manage.

So please, from a 20-year CISO who understands its better to design secure systems than to spend resources on compensating for the lapses, refine the downgrade flags to allow Fable to be used to design and write the security control systems of the applications.

Environment Info

• Platform: darwin
• Terminal: ghostty
• Version: 2.1.172
• Feedback ID: de3eb231-5a3f-44e7-abfb-05a47d502a1a

Errors

[]