Think small! Getting too ambitious will guarantee that you will fail.
Find ways to cut corners but do so with style. For example, with Anna, I'm not much of an artist, so to save time, I traced the main character over someone else's 3d render and made all the rest of the graphics in greyscale using extensive copy/pasting.
Focus on one or two unique things to make the game great, because you don't have time to throw too many WhizBangWow things in there. For Anna, I was mostly concentrating on the pseudo-3d presentation with keyboard control. The story is kind of pasted in (but wound up working out really well which surprised me at the end of the week) and the puzzles are super cliche, but that's all that I had time to come up with in the small period of time. The game succeeds, I think, on it's atmosphere and immersion, which are instilled largely by the pseudo-3d and direct control that were the focus of my development.