When designing a game, one of your main challenges is to expend as little wasted time as posible so you can focus on what is important: making fun.
You can read about agile methodologies elsewhere, so let' assume you grok that.
The first thing you need to do is define your main game loop, I.e. what does a player do in a game session until they get back to a starting point. In Super Mario, you start at the beginning of a level, move left to right, avoid obstacles and explore for puzzles and hidden items, and reach the end, where you receive a reward. In a racing game, you may select a car, type of transmission, start a race, pass cars, etc. until you cross the finish line, get rated, etc. As simple as this sounds, many developers don't start this way. Block out your game loop.
Then block out your menu/options system. This tells you if your team knows what they are trying to do in a way that forces some hard decisions. For instance, do you want to allow a selectable difficulty? Can you select a level? You might be surprised to find out how far off your team is from your original vision, or maybe they have better ideas. I guarantee that a menu discussion up front will be interesting.
Then you have to decide whether you are going to need to make things look perfect or if you can use rough art for everything.
I have had the most luck and the least wasted effort building projects with placeholder art first, and replacing it with final art only after it is playable and fun. This generally only works when you have control of the money, or the trust of the money people. You see, many people cannot "see through" the lack of final art to understand what it will look like once it is finished. Tread lightly there. You need to know the dangers. This can be mitigated by creating a small area that is final enough, so you can use it to show the money people.
Gotta run, unfortunately. Next time we'll talk about creating test areas for your features and more overall planning...