Reworking things a bit based on some excellent feedback.  I will be using an 8 bit ALU based on @roelh 's 4 bit ALU.  Instead of having hardwired to the AB, XY registers, it will have an input and output register.  It will then use the data-bus and it's input register for inputs and it's output can go anywhere one cycle later.

I have been thinking about addressing modes for instructions.  I think I will keep it to direct, register, and immediate.  I have been thinking about offset addressing, but with only an 8 bit ALU I will run into page overrun problems.  Better to do this in software I think.  I don't think it's worth building a 16 ALU, as this would introduce a whole new set of problems in this new design, and I have to remember to try and keep it simple. 

I've updated my block diagram, working on the new instruction set.  These are both a work in progress and very flexible.

https://docs.google.com/drawings/d/1BW84zwO5effvYcdKjxzyki9cuvuHsM98z6nAamw4iTM/edit?usp=sharing

https://docs.google.com/spreadsheets/d/1kbxYGDeIOszgvjPYRdes-bELOBlojT_tjxe_RD9MVUQ/edit?usp=sharing