Sorry, I misread your post. The model is linear so the IRFs must be symetric (with respect to the abscissa when you change the sign). So I suppose it is safe to simply put a minus sign before the curves characterizing the posterior IRFs.
Best,
Stéphane.