Code for cross lagged model with moderators

Question

I want to find out which relation exists between Attractiveness (X) and self-worth (Y) and use R for it. This relation should be moderated by age (A), gender (G) and social status (S). Because I want to find out, which direction the relation between X and Y has, I want to use a cross lagged model. The data is longitudinal and I have two waves in which they were collected. The moderation of A, G and S should be in the relation between X_t1 (attractiveness in the first wave) and Y_t1 (self-worth in the first wave) and between X_t2 and Y_t2. I expect a positive correlation between attractiveness and self-worth. How could the code look like?

I used a Cross-lagged-Panel-Model before. It looked like this, but didn't give me useful results (with u being the impulses):

model1tB0 <- '
  # regressions
    X_t2 ~ X_t1 + u_xt1 + Y_t1 + u_yt1 + A + G + S + Y_t1 * A + Y_t1 * G + Y_t1 * S + u_xt2
    Y_t2 ~ Y_t1 + u_yt1 + X_t1 + u_xt1 + A + G + S + X_t1 * A + X_t1 * G + X_t1 * S + u_yt2
  
  # impulses
    u_xt1 =~ X_t1
    X_t1 ~~ 0*X_t1
    u_xt2 =~ X_t2
    X_t2 ~~ 0*X_t2
    u_yt1 =~ Y_t1
    Y_t1 ~~ 0*Y_t1
    u_yt2 =~ Y_t2
    Y_t2 ~~ 0*Y_t2
  
  # co-movements
    u_xt1 ~~ u_yt1
    u_xt2 ~~ u_yt2
  
  # restrictions
    u_xt1 ~~ 0*u_xt2 + 0*u_yt2
    u_xt2 ~~ 0*u_yt1
    u_yt1 ~~ 0*u_yt2
'
fit0 <- sem(model1tB0, data = datenB)
summary(fit0, standardized = TRUE, fit.measures = TRUE)
'

The problem with this model was, that it was underidentified. With the impulses it was too complex and should be used for datasets with more waves. This lead to results that could not be interpreted. I want to get rid of the impulses to make it more simple. Furthermore my moderation variables a, g and s moderate the relation between X_t1 and Y_t2 and Y_t1 and X_t2 and not between X_t1 and Y_t1 and X_t2 and Y_t2. How do I change that? And which co-movements do I have to specify for my moderators?