Although design including code design moves forward in time, as soon as those are complete, you should ideally jump mentally into the future, starting with how the system will be supported and maintained after it is live and work backwards to the last relevant task which might be deployment of a given release into production. Does git cover it or do you need additional tools such as puppet? Will there be configuration files and installation scripts? What will the UAT and test environments be like and how will they be provisioned from your repository? How will you create the configuration data that distinguishes between say production, test and development environments and what tools will use that data? Only when you know that can you wind the clock back to the end of the development cycle where you can now make a list of deliverables by type, from your code design and implementation plan. After that you can jump back to where you are now in reality, design your git repository using your own correct rule (only push what is custom-built otherwise configure it to be installed wherever its needed) and you are good to go forward.