Serve GR00T Policy

Start one local, model-backed GR00T inference server and finish with a verified endpoint for an Arena Gr00tRemoteClosedloopPolicy client. Keep Experiment execution, artifacts, training, OSMO, and other policy providers outside this skill.

Read the checked-out workflow

Before acting, read:

• docs/pages/quickstart/running_a_real_policy/gr00t.rst for the maintained local workflow, hardware notes, checkpoint, embodiment, and endpoint.
• The server script referenced by that document for its current arguments and readiness message.
• isaaclab_arena_gr00t/utils/wait_for_gr00t_server.py for the protocol-level readiness probe.

Treat the current checkout as the source of truth. The maintained documentation currently launches from submodules/Isaac-GR00T, but explicitly marks that location as transitional. If the docs name a separate GR00T checkout after the refactor, follow them rather than recreating the removed submodule. Report any mismatch between this skill, the docs, and the server script.