Ok [Feature Request] Whitelist or exempt epidemiological modeling code from safety filtering

Bug Description
Re: Request ID: req_011CdgUjAgJrpHAp5JiEBKpc . This is an epidemiological model that's being thwacked by Opus 5's internals? I mean I saw it with Fable - which was also a mistake. And now Opus 5?

This is statistical research, not gain of function testing! Please help me understand why this happening as the model, while it presents an ID, does NOT present a specific reason for balking. And yes, I'm aware that the infosec minded do this purposefully, but I think we can afford at least a bit more transparency.

I've been working on a set of models and frontend visualization to examine the spatiotemporal patterns of several infectious diseases using historical data in a novel manner. This is part of a larger component of my research that's being severely protracted given that I'm using an interactive coding assistant to help with this work. I engineer and make the decisions on a spec and implementation plan. I let CC implement code that I'm not working on. I review/amend/edit. While the agent is coding one part, I'm working on another. Then short length code-reivews. Lather, rinse, repeat.

Fable displayed this behavior first, now all of a sudden Opus 5 blindsides me, and it's getting old quickly. And what ever happened to Claude for Science? Guess spatiotemporal modeling wasn't on the list? Seriously folks, you built the coding assistant which can be beneficial for single researchers such as myself who have no one else to help. But with the new pressures of researchers compressing timelines/milestones, this is quickly becoming the quick route to the next hellscape. Please roll out a fix for this or at the very least give me/us a process for making a request for whitelisting this kind of work.

Thank you.
Environment Info

• Platform: linux
• Terminal: cursor
• Version: 2.1.220
• Feedback ID: 69633bf9-49b2-42e2-b81e-4c3fb156af5f

Errors

[]