I'd say put all binaries in PATH, which appear to be useful to be started without ghidra in the first place.

more changes pending: I would like to attach or make a new PR with plugins functionality in the next day or two, but first I want to learn to use scm properly once and for all. "soon".

I found tig easier to use then plain git.

I have not abandoned this, it has just been more difficult than I expected and I've been otherwise busy.
Someone helped me by cannibalizing the plugin system from idris, I have my own implementation attempting to use makeScope, that doesnt work yet. I'm blocked on my poor understanding (and confusion stacked on confusion) of how makeScope and related things work.

The plugin system allows using plugins from an in-tree plugin specification attrset, as well as using overrides to attach additional plugins that may not be in-tree.

Here's an example (dump) of how I'm currently using it:

with (import ./nixpkgs.nix); #pinning

let
  gh_path = /home/path/to/another-attempt.nix;
  plugin_path = ./plugin.nix;

  #just consume the args, the original isnt an overlay (doesnt need to reference self/super
  mkOverlay = set: self: super: set;

  overrides = mkOverlay { MyPlugin = (import plugin_path); };
  ghidraPlugins = (callPackage gh_path { inherit overrides; });

  ghidra = (ghidraPlugins.withPlugins (p: with p; [ ghidra-scala-loader MyPlugin ]));
  eclipse = eclipses.eclipseWithPlugins { eclipse = eclipses.eclipse-java; plugins = [ ghidra.ghidraDev ]; };
in
  (symlinkJoin {
    name = "shell";
    buildInputs = [
      ghidra
      eclipse
      ];
    paths = [
      ghidra
      eclipse
      ];
    }) // { inherit ghidra; }

This is superseded by #60664 .