OK i can't focus on work and keep looking at this repo.
So after every "subagent" runs, claude code creates another "agent" to check on whether the first "agent" did the thing it was supposed to. I don't know about you but i smell a bit of a problem, if you can't trust whether one "agent" with a v...