The graphics and music can be solved later; you can use some temporary graphics while you work on the program.

It has the advantage that if you don't succeed in making the game -- or if your ideas gradually change so much that many of the original ideas will not be used in the final version -- you don't have to pay for things you don't use. Also, when the program is ready, it is easier to explain how the graphics will be used, and you can test it immediately. After playing the game you will also know whether some things should be made bigger or smaller, which would again have some impact on graphics.

(With regards to code, I am fan of reinventing the wheels, because it feels to me that I only need to reinvent each wheel once, but would probably need to learn new toolkits again and again just to find that some important functionality is missing or full of bugs. But this is not a recommendation, just my specific experience which may be not typical; and the experience is not with RPG makers.)

Could you be more specific about the RPG? There are many options: isometric or 3d, turn-based or real-time... Will it have a story, or just a lot of random fights? If the story, will there be branches and multiple endings? How complicated it the system of magic or whatever (is it enough just to attack and expect that a higher-level character automatically does higher damage, or is it necessary to use a proper action to solve the specific situation)? One character or a party? Is the game over when one of your character dies, or do you just recruit a new one? Etc. -- Think about how your choices fit together. For example, if the world is automatically generated or you replace dead members with new recruits, how will you keep a meaningful story?