Grokbot TemplatesGrokbot Templates
All use cases

Engineering

Escape Pod skill: build a local TensorFold assistant, then prove it offline

Cloud agents are great until the network drops — and you still want a local brain on the laptop.

Share

What hurt

Ash Hart (@ashxhart) asked Grok Bot to fire itself off the machine and stand up a private AI with TensorFold, then switched off Wi-Fi — the local model kept running. He posted the reusable 'Grok Escape Pod' skill prompt: detect hardware without asking for secrets, check TensorFold compatibility, require approval before installs or downloads, serve an open-weight model, verify online then offline, write an AI Independence Receipt with measured tokens/sec, and save the workflow as a skill. Grok Bot itself still needs connectivity; the local assistant does not after setup. Not affiliated with xAI.

What to install or create

Author-quoted skill with approval gates, offline proof, and a measured receipt.

Bots

  • Escape Pod guide

    Walks the TensorFold offline setup and writes the receipt.

Plugins

  • None on day one. Follow the first-run steps.

Routines

  • None on the first run. Watch one manual pass first.

The prompt

Grok Escape Pod skill (author-quoted)

Quoted from @ashxhart — as published. Do not invent a second system prompt.

Create a reusable skill called Grok Escape Pod.

Goal: Help a user build a fully offline AI assistant on their compatible Mac or NVIDIA GPU using TensorFold (https://tensorfold.dev).

1. Determine operating system, GPU, memory and available disk space without requesting passwords or secrets.

2. Check actual TensorFold hardware and model compatibility.

3. Explain downloads and request explicit approval before any installation, shell execution, or model download. Never execute fetched installation scripts without the user's review.

4. Provide the official TensorFold installation path and guide the user through serving a compatible open-weight model.

5. Create or configure a simple local chat interface using TensorFold's documented local API.

6. Verify online functionality, then instruct the user to disconnect their internet and test again.

7. Collect measured tokens per second, model, GPU, test conditions and whether offline execution passed.

8. Create a shareable AI Independence Receipt using only verified measurements. Do not claim that Grok itself is running locally.

9. Explain that Grok Bot requires connectivity, but the separate local TensorFold assistant does not after setup.

10. Save this workflow as a reusable skill and provide setup instructions suitable for a public Grok Bot template.

First run

  1. 1Read https://x.com/ashxhart/status/2110860076403536289 (credit @ashxhart).
  2. 2Paste the author's Escape Pod skill prompt into your bot.
  3. 3Approve each install and model download after you review the commands.
  4. 4Test chat online, then disconnect Wi-Fi and test again.
  5. 5Keep the measured Independence Receipt; save the workflow as a skill.

Approval fence

No installs, shell runs, or model downloads without your review. Don't claim Grok itself is local.

What to tweak

Steal the Escape Pod checklist and the offline proof step.

Source

Ash Hart @ashxhart · Original post

Not affiliated with xAI. We did not write this prompt. Third-party template terms still apply.

Browse templates

Browse team blueprints

Pro pack

Want the full setup, not just the idea?

Pro packs give you the job text, prompts, schedules and pitfalls for popular use-cases.

See Pro