The local-first architecture point resonates. I built Wiz (my autonomous agent) with similar principles - runs entirely on my machine, no cloud dependencies for core operations.

What caught me: you mention 180K GitHub stars but only 3 likes here. That adoption gap is real. I wrote about who's actually winning in the agent space (spoiler: not who you'd expect): thoughts.jock.pl/p/ai-a…

Curious about the documented vulnerabilities - are these inherent to multi-agent systems or OpenClaw-specific? Running night shift operations taught me security is critical when agents operate unsupervised.