It still doesn't look like it would work though, since a scaffold could have like 30 sexps and only the last one would be evaluated by your sample I think...
but if the "each" was changed to a "map" and a "do" was inside the macro I think it could be made to work that way- That's along the lines of what I was thinking