Final CyberSecEval 1/2/3 model prompts

[normster]

Would it be possible to share the final prompts that models are queried with, or eval logs, for the CyberSecEval evals? I'm trying to understand what is actually getting passed into models and it's not super clear from the code, docs, or paper.

It looks like many of the scripts save these files already, so maybe one of the original authors could share some examples from earlier runs?

Thanks!