Just to clarify. Could you show which directory you are in when you are using CLI, and what command you are trying to run?
Also, could you check your PATH, whether the Chia folder is there.
By the way, it looks like the Chia original code came from Linux side, as such some things are not really translating well into Windows. One of those is that the file names / paths are case sensitive in Linux. So, having Chia as the GUI starter and chia as the main process is OK under Linux, but is rather nonsense on Windows side. As Chia.exe (capital C) is only meant to be run to start UI, you can safely rename it to ChiaGUI.exe or whatever you like. This way, you will take it out of the way from the eventual path (whether explicit or implicit). Whatever initialization should be done was done when you run the setup, so there is no need on the harvester side to have Chia.exe around after that.
Open the ~/.chia/mainnet/config/config.yaml file in each harvester, and enter your main machine’s IP address in the remote harvester’s farmer_peer section (NOT full_node).
For example, this section would look like:
host: 192.168.2.13 <----this is my farmer static IP
.\chia start -r harvester
If you want to plot on that same machine I would use madmax plotmanager from CLI