estimated variable as dependent variable get robust standard errors


I have a problem with the calculation of standard errors when using a dependent variable in a regression model that was previously estimated from another model. I would like to use some kind of sandwich estimator using inside the standard errors from the first step model. Can anyone help me?