I have decided I wanted to understand one basic opening against 1. e4 and one against 1. d4 as black. After watching a few videos on multiple openings and systems, I decided I wanted to try and pick up the Caro-Kann. I was wondering if it's sensible to build an opening repertoire there. More specifically, is it possible to do it without chessbase? Is the free stockfish analysis reliable? What steps should I take to ensure my analysis isn't flawed?
Create a private Lichess study and add your game references there. As you keep adding more of your learnings and research, the study will develop into a full repertoire. You can also export as png, so you can always switch to other software if you find a better option.
I want to give you a small suggestion related to B.Swan's comment:
The engine will not lie to you, but if a position is equal to SF11, is it equal to you too? Very often not. Then it would not be reliable. But you will have this problem with any engine. In any variation you analyse, you should not be overly concerned about ultra soundness (oh [no] the position is +0.2! where have I gone wrong), but rather about whether it is clear to you how the game should proceed and whether you can handle it (and of course whether you like it..)
I think it is best if you build your study based on your understanding first, before checking with an engine. The reason is as B.Swan mentioned, namely that what an engine can cope with is often not what you can cope with. So what you want is not the best move based on very deep tactics, but the best move based on what you can analyze. No doubt, this means that you will not play as well as the engine, but because of that you may play better than if you tried to mimic the engine. The reason to check with the engine is not to ensure that you have found the best move, but rather to ensure that you have not from any position in your study chosen a move that can lead via deep tactics by the opponent to a disaster for you!
In other words, you use your own skill to find your own moves, but use the engine to find opponent responses to those moves to make sure they are not mistakes or blunders under best play. Such an asymmetric use of the engine is more likely to improve your play than to attempt to follow only what the engine prefers best, because humans simply do not have the same high-fidelity storage capacity as computers.
There’s no need to buy chessbase (although I’m not arguing against it if you want to do so). Chess.com actually has an excellent database of masters games available to free users. The Lichess opening book is compiled of 2 million+ master games but only a small number of these are available as a pgn for you to load in the Lichess analysis board. I think chesscom is the better resource for this.
The flip-side of this is that the chesscom opening book is crippled for free users. I’d recommend using the chesscom games database and the Lichess opening book and between the two, you pretty much have chessbase.