consider strict separation of code and data

For my own brew stuff, I generate data files of my installed stuff via this shell function:

save_brew () {
    brew leaves > ~/projects/dotfiles/brew_leaves.txt
    brew cask list > ~/projects/dotfiles/brew_cask_list.txt
}

So I can just install stuff normally via brew and if I decide I want to keep it I can save_brew. I can install with:

brewify() {
  cat "${DOTFILES}/brew_leaves.txt" | xargs brew install
}

I think you'll find maintaining this easier if you move toward keeping code/logic separated from pure data of installed inventory.

Random side note that I like your notion of using golang for a new version of this.

Start by inspecting the repository's current bootstrap entry points and how it handles installed Homebrew inventory; the issue names the save_brew and brewify shell functions as the proposed reference. Clarify the intended code/data boundary and whether the possible Go rewrite is in scope. Done should mean the inventory is maintained as data separately from the installation logic.