[Feature Request] Differentiate security-enhancing code modifications from security-degrading ones in safeguards

Bug Description
CRITICAL (no not really im just flagging this): The fable safeguards are well implemented. But its particularily frustrating when it blocks attempts to increase code and agent security. I think a focused approach should be to separate security enhancement from security degredation, and aim to allow more of the first and less of the second. Though I understand this separation is sometimes diffuse, being able to use the most advanced models for security and not attacking, is itself a security layer, like an inverse roman shield formation saying "defense is the best offense"

Environment Info

• Platform: darwin
• Terminal: vscode
• Version: 2.1.173
• Feedback ID: cfab367d-f08e-4d01-a43b-73a71ef934d9

Errors

[]