There have been a lot of questions from new users that have been put on hold for being too broad. Most of these, thankfully, have a comment from another user referencing the help pages. However, if a new user understands the scope of the site, but is still struggling to formulate their question, I want to be able to refer them to a solid "good question with good answer" that demonstrates the expectations. Are there any questions that are commonly used for this purpose?
A while back we had a series of case studies where some users showed off their top-voted questions and we all gave our opinions on if the questions were good or not. I wouldn't say we ever found a specific list of 'good' questions, but some users did stand out as asking really good questions. For instance, I know I remember having very few issues with James's questions.
While this may not be what you wanted, the idea of the series was to better understand what makes a question 'good', or 'on-topic', or otherwise 'not something we should close'. While we never did actually come to a consensus, I think the opinions presented in the responses to each post may help new users understand potential dangers, and perhaps give them a few examples that they can model their own questions off of.
I'm surprised this hasn't been floated yet: One Year of Building Worlds.
They're specifically what you're asking for. Questions and/or answers nominated by users as examples of what's best around here, with some commentary (both in-post and in-comment) as to what makes them so good. For my part, some of my favorite hours on this site have been reading others' answers to this meta question.
I think it's a more-approachable list than the case studies: there we've got lots of good discussion of risk factors, axes of critique, and other terms of art we've devised that may be noise to a new(ish) user.
Another alternative, drawing on bilbo_pingouin's suggestion in the comments, is the aggregate list of questions that have been posted after going through the Sandbox. The list is updated manually, but seems reasonably complete.
These are questions that have been proposed by community members seeking feedback on them, vetted by the community on meta for obvious problems, and only after that been posted on the main site as actual questions to be answered.
The process obviously isn't foolproof, because nothing is foolproof (the universe just keeps inventing better fools) but it does ensure a minimum level of quality by the time the question hits the main site, and the activity seen through the sandbox also helps ensure a certain level of activity on the question also once it hits the main site.
The aggregate list includes questions on a large number of topics which means most users should be able to find a question in the list that is about a topic at least similar to what they want to ask about.
The sandbox is also available for fairly new users (a nominal amount of reputation is needed to participate on meta; one upvote on any post on the Worldbuilding main site, or the association bonus, is enough) and thus may also be available to newcomers to the site for receiving feedback on their questions.
I would generally agree with @DaaaahWhoosh, part of the problem with your request (its a good idea mind you) is that the range of topics and question types is so far reaching on world building that a single (or even half dozen) good question is hard to use as a broadly applicable model.
The case study series is probably the best suggestion I can think of as well.