Personally the way I do game jams is to have big ideas and then from there start saying no to myself! The reasons can vary but my most common ones are that it takes too long to implement so I say no or that the core of the game is still there if I don't do this feature. Once you have your core you want to make the system as unsophisticated as possible, make it simple and easy to implement while achieving the same goal as if you had developed something more complex.
The game I submitted for this jam (called From Beyond the Wall) had a good amount of "no don't implement that". My plan starting out is to have a much longer level and a real time lighting system but after starting to develop it and seeing how much time I have left I just said no to the two of those and went with a shorter level and the light in my game is basically just a circle collider that turns on or off every few seconds to determine when the monster can or can't go through. Speaking of the monster, it's literally just an object that moves towards the player with minimal AI that interacts with the 'light'.
As for planning I can't help you there, my workflow is too sporadic and chaotic :(