My new robot is leveraging AgenticROS to connect AI Agents (OpenClaw, Claude, Gemini, Codex, and Hermes) to ROS2 to give AI control over its own body (RealSense + Wheels + Legs + Arms etc). AgenticROS also now has a skills marketplace available at https://skills.agenticros.com! It’s open sourced under Apache 2 and I would love your feedback!! 
Hi Chriss,
Watched the video, cool looking robot!
The eyes looking around give it real presence, and driving the whole thing across five different agents through one tool surface is genuinely impressive. The tall single-mast build on a small wheeled base is a fun, hard form factor to work with too.
I build in exactly this area, a few things I got curious about watching it move (the little reverse-and-turn after the eyes come up): on a tall base like that, what’s handling the body side of safety? A couple of specifics I’d love to know: is there anything independent of the agent that can refuse or stop a motion command based on the robot’s actual state (tilt, base velocity, an obstacle in the direction of travel), rather than trusting the model’s output? And is that startup movement a homing/calibration pass or the agent acting on its boot context? are you capping the robot’s move speed to preventing from tipping over (for example) and if so, how are you enforcing that?
Either way, love seeing this out in the open. Curious how you’re thinking about that layer as more skills land in the marketplace.
Cheers!
Julio
