Refactoring a multi-agent supervisor flow where OpenClaw and Hermes agents collaborate for self-impr

The #1 question I keep getting is "Why not just use Hermes for everything?"

Since Hermes decided to swallow OpenClaw's critical workflow without my permission, I migrated everything over.

But now we have to refactor the multi-agent supervisor flow by:

- Turning OpenClaw into the supervisor agent
- Building a Hermes supervisor agent for Hermes workflows
- Make them both talk to each other to solve problems and make improvements

Luckily, @NousResearch deployed multi-agents today for Hermes in v0.5.0

If anyone's interested in my multi-agent setup designed for mass self-improvement, I'll share it here.