Train positions where you have to make a meaningful choice. A useful position may contain a pawn break, a development decision or a tactical threat that distinguishes two variations. Starting from that point tests recognition directly. You cannot rely entirely on the rhythm of replaying the opening from move one, so weak understanding becomes easier to notice.
Before moving, describe the position in three short observations: whose turn it is, what the opponent threatens, and what your side wants to achieve. In the Caro-Kann diagram, Black has exchanged the d-pawn for White's e-pawn and White's knight sits on e4. Developing the light-squared bishop actively is one important option. Explain why the square is available before looking at a continuation.
Use both recall and analysis exercises. A recall exercise asks for the move in your selected line. An analysis exercise may have several reasonable candidates and asks you to compare them. The daily trainer tests repertoire recall; an alternative legal move being rejected there does not necessarily mean it is a chess blunder. Investigate plausible alternatives separately when reviewing your games.
After answering, inspect the feedback and revisit the position after a delay. If you used a hint, make the next attempt without it. Vary the starting points and include positions from different selected openings. Stop adding new material when you can no longer explain the moves: a smaller set of understood decisions is easier to apply in actual games.
Try this at the board
In the diagram, explain the idea of ...Bf5 and compare it with developing a knight first.