[Feature Request] Add configurable safety threshold to reduce false positive model access denials

Question

**Bug Description**
Create a skill for Fable such that it can intelligently avoid false positives; maybe warn the user first before denying access to the model.  This is not an attempt at reverse engineering, just any way of reducing the constant false positives as I try to code a simple MVP for an internal product.

**Environment Info**
- Platform: linux
- Terminal: kitty
- Version: 2.1.206
- Feedback ID: b7f1b4e1-4702-4687-9861-ad5e35de7037

**Errors**
```json
[]
```