I assume you are speaking of ACA software.
In modern versions of ACA software (ACA/Web v1 and later, from about 1998 and later), you cannot exclude some of the attributes from the "Priors" section for respondents. (Prior to 1998, they could only be excluded due to asking respondents to indicate that certain attributes were "unacceptable"). In ACA/Web v1 and later, respondents either get a "priors" section (with all attributes) or not. If you include the priors section, then all attributes need to be included at least in the Importance phase. If you specify that some attributes have "a priori" level orders, then those attributes are dropped from the Ratings.
I recommend users strongly consider whether the entire Priors section can be skipped in ACA. If you have large enough sample size, and are using ACA/HB estimation of the utilities, then it is possible to do this. Respondents then only get the Pairs section (the "true" conjoint portion of the interview). Past evidence suggests that if you have large enough sample sizes, this actually can work better for getting utilities that are predictive of holdout choice behavior.
Please read the white paper we wrote on dropping the Pairs section in ACA if you are interested in dropping the Importances and Ratings sections altogether: https://www.sawtoothsoftware.com/download/techpap/omitimp.pdf