this might sound counter-productive but it changed how i ship with AI agents... as you may know, Hermes can control Claude Code (it's also a great option to use Anthropic models from Hermes) but i also never let Hermes start from zero on big projects before i hand it off, i set up the Claude Code environment myself: > the initial repo structure > the first claude .md > the rough plan then i let Hermes build its own skills, populate the directory, do the work and whenever i need to work on…