You _could_, but I don't recommend it. The "decode" box in the diagram really does not mean that there is just one chip for the decoding. The decode box and all the d circles together mean, "the stuff you need to build to get the CPU to work."
I recommend picking a starting point, and moving from there. For example, start with the A register, or PC.
One of the problems with making an "instruction decoder" chip is that the interface is very broad -- lots of inputs and outputs that each need a connection in the HDL line that uses the decoder. I found it better to just put the code directly in CPU.hdl.
I do have it all together in a block of code labeled "// Instruction decode".
When I implemented the Hack CPU in Logisim, I did make an InstructionDecoder part. It made for a much cleaner and easier to read circuit given the schematic-based interface Logisim uses.