I agree. However as you said, you can consider code agents like human contributors. My feedback is that an army of Claude Code instances with a strict CLAUDE.md file is more rigorous than a human team.