Just did a quick test with an old image. I saw many examples but used
this one
to see how it does on my usually too dark images.
I think it is fairly easy to either analyse the guider images or the real images and do a prediction of cloud coverage once the star-count has dropped significantly.