But if you've done: "$(fgcolor red)red text$(fgcolor green "green text") what color is this?"... then it would have to model a stack on top of the actual underl

But if you've done: "$(fgcolor red)red text$(fgcolor green "green text") what color is this?"... then it would have to model a stack on top of the actual underlying implementation, which does not have any concept of "closing" a sequence to pop the colour back to the previous colour. It could automatically emit the \x1b[0m, but it would be making assumptions about nesting.