Use ideas that you have from other openings and combine them with your own to make your own ideas. Play games with your custom countergambit, make your own theory, get it analyzed by other players, improve it, go through many iterations, keep building.
Its not going to be easy to make a whole custom opening line, especially a gambit for black, but it will be fun.
I might come up with a concept too, as its quite rare to have such an attacking opening for black (aside from the few already made) nonetheless a custom original one.
Please take my words with a grain of salt though, as I probably dont know what I am talking about.
Before I start, I want to make it very clear that, yes, while everyone knows the same old tune, a comfortable attack is better than any gambit or what have you, and everyone's probably going to think that trying to play a gambit is rather pointless regardless of level of play, and "I want to invent a new gambit" is like, you've heard it a billion times, I get that, BUT, the actual reason why I'm trying to invent a gambit as opposed to some form of defence is only because I wanted something that would actually suit my style of play, so it's not like some kind of "egg gambit" or "Botez gambit" or "alphabet gambit" is even going to suffice here. It has to really check all of the boxes, and really, the more important part of this journey is to better understand the process of how an opening becomes registered on lichess and chesscom and FIDE and how I can get the privilege to name it and so on.
1. Normally when you hear an "attack" opening it's usually an opening for white. What I want to invent is not that. It's going to be an opening that is still called an "attack", but much like the Marshall Attack, I'm trying to invent an opening for the BLACK player to launch a counterattack with. Some sort of "countergambit" as if it were, so that for whoever winds up with the black pieces, if they were to play this, they wouldn't be in a position where they're defending anymore. It's not that I want chess to be "interesting" again or anything. I do want players to win, but what I feel is that the top levels of competitive chess feel dragged on for longer than they could have, simply because very often, players focus more on playing defensively until their opponents make the one fatal mistake and then they proceed to push, when, if even the players with the black pieces just went off counterattacking instead while still playing pretty sharp, tactical, and calm, I have a hunch that these countergambit tactics could be far more efficient, and the games don't have to be dragged on for longer than they have to. It's not a matter of players taking too much time to think, it's a matter of coming up with something that can win the game in fewer moves so that even the players can be home in time for dinner (I mean that in a serious way).
2. This is not meant to be some fun opening for people to try. We want this counterattacking opening to be playable/winnable and foolproof at the very top levels of competitive play. We don't want it to just be interesting. It's not even about being interesting. We want it to be efficient, which means we want this countergambit to be really effective at ending the game in fewer moves. We want it to be foolproof, which means we want this countergambit to spring out of a very common line that white would play, e.g. we gambit the pawn within move 2 or 3 where the white player will have played the most commonly seen moves as if this were some kind of anti-London, or we see some kind of main opening idea often enough such as Scotch or Italian and then we proceed to gambit a pawn from there. Also, we definitely want this to be a serious line, which means we want it to be something that's competitive enough that even the top level players would consider playing.
3. This is more so about learning the actual process of how a new opening/gambit officially gets registered on chesscom and other sites, very much like how Frank James Marshall secretly studied his attack opening for 10+ years through all of the potential research and prep he had been doing all that time. Obviously we would probably want to do this on our own end via downloading a database for a chess engine to analyse and then give it an external NNUE so it wouldn't play the same defensive style over and over (How does one install an external NNUE onto Stockfish for Arena GUI? I'm trying to make something like Grasshopper work for example, am I missing something? And how do I give it a database? Do I just drop it into the folder?) and probably memorising all of the potential lines and figuring out the "main line" of sorts and mentioning a bunch of variations and so on and actually learning how a new opening becomes "invented/discovered" as it were. It's about learning all of this process, not just trying to figure out a better way to attack with the black pieces. The journey and context is what matters, not the result. I want to personally re-live what it was like for Frank to come up with his attack opening.
TL;DR Finding a really effective pawn sacrifice as black would be nice, but what I'm REALLY here for is to learn what it was like for Frank James Marshall to invent his Marshall Attack and how it got officially registered on lichess chesscom and so on, and I want to learn how to do this myself and re-live his journey and come up with something of my own, to learn what it was like.