You can estimate this model within the framework of path analysis / structural equation modeling. The moderated effects can be represented and tested by including interaction (product) terms as in moderated regression analysis (e.g., computing and including a product variable LUX*OKU to represent the potential interaction between the two conditions).

The interactions between the binary condition variables (LUX, OKU) and latent variables (e.g., KS*LUX) are more difficult to model if the latent variables are represented with multiple indicators (i.e., if there is a measurement model with multiple observed variables for each latent factor). In that situation, latent moderated structural equation modeling (LMS; Klein & Moosbrugger, 2000) or the so-called product-indicator approach may be used to model the latent-by-observed interactions (both would be complicated in your case though given the presence of multiple interaction effects in the model). To my knowledge, LMS is not currently available in AMOS (it is available in Mplus), but you could potentially use the product-indicator approach. See, e.g.,

https://psycnet.apa.org/buy/2004-17801-001

https://www.tandfonline.com/doi/abs/10.1207/S15328007SEM0901_1?src=recsys

Klein, A., & Moosbrugger, H. (2000). Maximum likelihood estimation of latent interaction effects with the LMS method. Psychometrika, 65(4), 457–474. https://doi.org/10.1007/BF02296338

An easier way would be to represent the KS, PS, ZSM, and BSS variables by observed (sum or mean) scores (rather than by multiple indicators). In that case, you would not have to deal with the complications of modeling latent interaction effects. When all variables are observed variables in your data set, you can simply compute and include manifest product terms to model the interaction/moderator effects.

I think it would be better to use SmartPLS software instead of AMOS as it would give you more control in my opinion.