Wouldn't completing a list of goals contradict "not ask leading questions"? Why not just give the app to the user and tell him/her to do his job with it? Instead of telling him what to do.
In the end, the app is supposed to help him accomplish what they want, not what you want them to do.
No, it's not a contradiction to have an agenda in mind and to also try not to ask leading questions.
An example of a leading question in this context would be like asking them "do you see the button labeled 'learn more' to your right?", which really doesn't test anything other than the fact that they're literate and their eyes work.
Compare this to asking them "now what do you see on the page and what do you want to do?" Maybe they'll want more information and hit the learn more button, maybe they want info and don't see it, or maybe they want to do something else entirely.
And sure, you can just give the app to the user and see what they do, without writing up goals in advance. That's the unscripted method that Mark Hurst popularized and I linked to. Here's an even longer article by him on the same method:
But that's not appropriate for every situation. Unscripted stuff is great for getting the lay of the land and uncovering as many problems as you can, or finding out if they'd use your app at all even if you got rid of all the obstacles. Often times you already know about specific usability issues, however, or you want to make sure they cover all the bases in an app with a lot of moving pieces.
In the end, the app is supposed to help him accomplish what they want, not what you want them to do.