Ciying AI Directory
AI DirectorySkills
中文
Ciying AI Directory

NVIDIA NemoClaw Getting Started in Codex for a Preflight Review

Prepare for NemoClaw onboarding by reviewing the right agent variant, prerequisites, host changes, and the first sandbox checkpoint before installing.

中文

Setup for Codex (NVIDIA skills)

Use the NVIDIA skill in Codex. The example is an offline preflight based on a fictional host. Live onboarding uses a supported operating environment and the current NVIDIA quickstart for your agent variant.

  1. Open the pinned NVIDIA skills repository and add the NVIDIA getting-started skill to Codex. For the historical revision, copy its whole source folder with references into the practice project's .agents/skills directory.
  2. Invoke $nemoclaw-user-get-started and choose the OpenClaw or Hermes guide for the intended test. Read the current NVIDIA prerequisites and quickstart before starting the runtime installation.
  3. For a future live test, review the installer and note any Docker, operating-system, gateway, or account changes it expects. Follow the documented onboarding for your chosen variant on a test host.

Example task

Ask Codex for a preflight table for a fictional Linux test host. It has no working container runtime, no provider selected, and no existing NemoClaw sandbox. Request a check of prerequisites, required decisions, and the future first-prompt verification step without running commands.

Expected result

The table marks the container runtime and provider choice as pending. It leaves the sandbox and first-prompt checks unverified and lists where the current NVIDIA documentation resolves each prerequisite.

Upstream documentation

Original SKILL.md on GitHub · Apache-2.0 license · Source reviewed

Frequently asked question

Does passing the preflight show that the agent is running?

The preflight checks prerequisites. Confirm a running sandbox and an actual response only after you complete the variant's documented onboarding on your test host.