I am running a two step selection regression in stata, where there also exists an endogenous variable. The outcome variable of interest is a binary variable. I am not able to understand which model to execute in stata

1) Following the theory from Woolridge (2010), I run a probit for the selection equation and calculate the IMR. Place it in the main IV 2sls regression. However, since a generated regressor is being used, I need to bootstrap the standard errors to get accurate estimates.
2) I use eprobit in stata. Do I need to bootstrap the standard errors in this case as well?
3) I use Murphy-Topel correction of variances in heckprobit. But as I was reading from Muro, Suarez and Zamura (2010), the endogenous variable is not continuous. So it is not possible to use scores in the predict command.

It would be really helpful if I could get any assistance on what method to apply.