Yes, you can compute variables (e.g., composite scores or means of items) in SPSS after conducting validity and reliability tests, and then use them directly in AMOS—but this is only appropriate if you're conducting path analysis or regression using observed variables, and not full structural equation modeling (SEM). If your goal is to model latent variables explicitly, account for measurement error, and assess both the measurement and structural models together (as in SEM), then you should not use computed scores. Instead, use the original items in AMOS, draw latent constructs, and connect them to their respective indicators. Using composite scores bypasses the measurement model, which is a core part of SEM.