This release is meant for internal use only. Run it on a lab machine or an internal network that you control, operated by people you trust. It is not built to be exposed to the internet, to untrusted users, or to more than one tenant.

Report a vulnerability to support@cheshirelabs.io. Please do not open a public issue for one.

orca start binds 127.0.0.1, and every route on it is open to every process running on that computer as any user. There is no API key and no per-request check. A caller that reaches the port can move an arm, dispense a reagent, abort a run, and run arbitrary Python as the user who started the daemon.

Workflows, methods and actions are Python source, and the daemon imports and runs them. POST /operations/insert-method, /insert-action, /replace-method and /replace-action accept source over the wire and execute it. The import and builtin restrictions in orca/runtime/code_injection.py catch a mistake, not an attacker, and they are not a sandbox. A deployment profile's computed expression is the same: it runs with the deployment's own reach.

So run the daemon on a machine only your operators use, do not run untrusted code as the same user, and do not expose the port through a tunnel, a proxy or a container port map.

This software drives real laboratory instruments. A command can move a robotic arm or dispense a reagent, so a mistaken or hostile call destroys samples and can damage hardware. Operator device writes default to LIVE, which is deliberate: an operator typing initialize means the real device.