Jetson Orin Nano Offline Setup

I have a Jetson Orin Nano Developer Kit that still needs its very first flash. I need a self-contained set of files—straight from NVIDIA—that will let me bring the board up without ever touching the public internet. After the board boots, I want to install and run OpenClaw together with a local copy of Llama so I can:

• keep the system completely air-gapped,
• pull live SCADA data, and
• interact with it through the Llama model or route commands through OpenClaw.

What I expect from you
1. A repeatable, step-by-step guide that starts with downloading the correct JetPack image, SDK Manager-ready packages, and all dependencies, then walks through flashing the Orin Nano from an offline host.
2. An offline installer (or clearly scripted process) for both OpenClaw and the Llama model version you recommend, verified to run locally with no external calls.
3. Configuration instructions or sample code that demonstrates how to ingest typical SCADA data (Modbus/TCP is fine as an example) and expose it to Llama/OpenClaw for local querying or control.
4. Confirmation procedures so I can verify the board stays isolated—no DNS lookups, NTP calls, or telemetry once it boots.

Additionally: The setup will eventually integrate two additional Nvidia 128GB computers and two air-gapped LLMs into a single connected system after completing the Nano setup.

Deliverables must be copy-paste ready and tested on the Jetson Orin Nano Developer Kit. If anything requires a specific Linux host toolchain or version of JetPack, note it explicitly.