RhinoClaw

Control Rhino 3D via AI agents. 72+ tools for geometry, transforms, booleans, PBR materials, Grasshopper automation, VisualARQ BIM objects, and viewport cont...

This skill appears to do what it claims — control Rhino via a RhinoClaw plugin — but it is powerful and should be used with care. Before installing or running it: (1) Only connect to Rhino hosts you trust (prefer 127.0.0.1 for local Rhino). The SKILL relies on a Rhino plugin that listens on TCP; exposing that port to LAN/Internet is risky. (2) Be aware the skill can execute arbitrary Python inside Rhino (execute_rhinoscript_python_code) — do not allow untrusted agents or people to invoke those flows. (3) The GH deploy flow copies files into your home directory and may run a local meta_gen.py subprocess; review the destination path and meta_gen.py script before running. (4) If you need higher assurance, open and review scripts/rhino_client.py and script_exec.py to confirm there is no hidden network exfiltration or calls to unexpected endpoints. If you want help reviewing any specific file (for example rhino_client.py), provide its content and I can point out any risky lines.

Type: OpenClaw Skill Name: rhinoclaw Version: 0.2.6 The RhinoClaw skill bundle provides extensive control over Rhino 3D, including high-risk capabilities such as arbitrary Python and RhinoScript execution (script_exec.py, ghscript.py), file system operations (files.py), and screen capture (viewport.py). While these features are aligned with the stated purpose of CAD and BIM automation, they represent a significant attack surface for remote code execution and unauthorized file access within the Rhino environment. Additionally, ghscript.py makes specific assumptions about the host's directory structure (~/projects/rhino-compute-platform), and rhino_client.py allows for log retrieval. No evidence of intentional malice, data exfiltration, or obfuscation was found, but the bundle's powerful primitives require high trust.