A game always starts out as nothing more than an idea. The idea can be derived from or even blatantly copied from one or more existing games, or can be almost entirely new and revolutionary, but the designer needs to have something in mind before starting to document and build it. Typically the first document to be written is a sort of brief pitch document. This can then be used to convince other people the idea is good enough to work on, and if approved (or if you think just you like it well enough in the case of solo development), expanded upon to create an outline of all the various parts of the game that will need to be created.
Regardless of the exact process a team or individual uses for early documentation, things naturally proceed from there to some form of early prototyping. A lot of ideas can sound good in writing, or may seem reasonable until you try implementing them. Prototyping is the first test of whether an idea can actually be fun, or if it is even possible to implement given the capabilities of the tools that are available.
While it might be tempting to just start making the final version of the game straight away, building a prototype first is almost always the best approach. The goal of a prototype is to get something working fast. This means you should cut every corner, throw out any requirements for scalability or maintainability, and just get something together that functionally works as quickly as you can. Art assets can be colored cubes, or free models downloaded from the internet, UI can be raw text with no style or fancy fonts, etc. Typically the uglier and less polished your prototype is the better, because that means you spent less time on it, which is less time wasted when you throw it all away.
So why even build a prototype if you’re going to trash it before you start on the actual game? Because the prototype is your proof of concept. Above all else, your prototype needs to be fun. If your game isn’t fun as a prototype, it almost definitely won’t be fun as a final product. The key goal of a prototype is to help you fail quickly. You might build a dozen prototypes before you find one you really like. If you just went ahead with your first idea and got all the polished code and final art together without prototyping, you might end up with a game you can’t actually finish, or that just isn’t very fun. A prototype shows not only that your design is actually possible, but that it is fun even in a basic state, which means it will be even more fun in it’s final version with a high level of polish.