If we're just collecting basic ways of detecting visibility, let me not forget:
opacity > 0.01; // probably more like .1 to actually be visible, but YMMV
And as to how to obtain attributes:
element.getAttribute(attributename);
So, in your example:
document.getElementById('snDealsPanel').getAttribute('visibility');
But wha? It doesn't work here. Look closer and you'll find that visibility is being updated not as an attribute on the element, but using the style
property. This is one of many problems with trying to do what you're doing. Among others: you can't guarantee that there's actually something to see in an element, just because its visibility, display, and opacity all have the correct values. It still might lack content, or it might lack a height and width. Another object might obscure it. For more detail, a quick Google search reveals this, and even includes a library to try solving the problem. (YMMV)
Check out the following, which are possible duplicates of this question, with excellent answers, including some insight from the mighty John Resig. However, your specific use-case is slightly different from the standard one, so I'll refrain from flagging:
(EDIT: OP SAYS HE'S SCRAPING PAGES, NOT CREATING THEM, SO BELOW ISN'T APPLICABLE)
A better option? Bind the visibility of elements to model properties and always make visibility contingent on that model, much as Angular does with ng-show. You can do that using any tool you want: Angular, plain JS, whatever. Better still, you can change the DOM implementation over time, but you'll always be able to read state from the model, instead of the DOM. Reading your truth from the DOM is Bad. And slow. Much better to check the model, and trust in your implementation to ensure that the DOM state reflects the model. (And use automated testing to confirm that assumption.)