agentleFS
NVIDIAKnown publisherNVIDIA/dgx-spark-playbooks/nvidia/station-ai-skills/assets/skills/dgx-station/SKILL.md

dgx-station

Skill1.5k starsChanged 2 months agoApache-2.0View on GitHub

Inspect and guide NVIDIA DGX Station GB300 development using the local dgx-assist CLI and pinned NVIDIA playbooks. Use for general Station platform questions, Software 1.0 or 2.0 compatibility, GB300 or RTX GPU selection, UUID ordering, mixed ATS/HMM coherency, CDMM, general containers, CDI, CUDA visibility, or vsloshd power-sloshing behavior. Do not use for vLLM or SGLang container selection or tuning, serving a named model, changing MIG, or troubleshooting a reported failure when the dedicated Station skill applies.

Trending

More talk lately.about the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 0 mentionsMar 2026: 0 mentionsApr 2026: 0 mentionsMay 2026: 0 mentionsJun 2026: 0 mentionsJul 2026: 0 mentionsAug 2026: 0 mentionsSep 2026: 0 mentionsOct 2026: 48 mentionsNov 2026: 2 mentions
Dec 2025Jun 2026Nov 2026
Mentions a month of the NVIDIA/dgx-spark-playbooks repository that our weekly crawl found. There is no adoption data for this file yet.

Who's talking about NVIDIA/dgx-spark-playbooks

The biggest 5 of 50 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 50

Talked about on GitHub (50 times).

On GitHub · 50

45 more on GitHub

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 7 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read references/software-compatibility.md.

    line 13
  2. Very specific:

    Before any platform action, require its named capability to be true.

    line 17
  3. Very specific:

    If version_specific_guidance is false or retrieval abstains, say no applicable version-specific passage was found and avoid inventing a …

    line 18
  4. Distinctive:

    Before any mutating action, display the exact command to be run and obtain explicit user approval immediately before executing it.

    line 23
  5. Very specific:

    Never assume an nvidia-smi index is a CUDA ordinal.

    line 24
  6. Distinctive:

    Use GPU UUIDs for every proposed launch.

    line 25
  7. Distinctive:

    Do not apply either legacy or Software 2.0 mixed-device behavior without checking the detected profile and observed ATS/HMM state.

    line 26
  8. Distinctive:

    Do not tell the user to install or operate Fabric Manager as a normal Station requirement.

    line 27
  9. Very specific:

    When mixed_coherency_service is true, inspect mixed-coherency-gpu-select.service, its generated environment, and container exposure …

    line 28
  10. Very specific:

    When dynamic_power_sloshing is true, inspect vsloshd.

    line 29
  11. Distinctive:

    Never install or change the driver, kernel, OS, firmware, or packages.

    line 30
  12. Distinctive:

    Never display credential values.

    line 31
  13. Distinctive:

    Read references/platform.md when interpreting coherency, container ordering, power, memory placement, or qualification evidence.

    line 33

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

DGX Station

Ground every Station-specific answer in current host evidence and a retrieved NVIDIA passage.

Workflow

  1. Run scripts/dgx-assist system inspect --json.
  2. Read references/software-compatibility.md. Use compatibility.profile_id, support_level, and capabilities; do not branch on a version prefix.
  3. Form a narrow playbook query from the user's question and run scripts/dgx-assist playbook search "<query>" --json.
  4. Apply the source-precedence rules in references/sources.md. Use only passages applicable to the detected profile. Cite their URL, heading, line span, source commit, and source-file digest.
  5. On the Software 1.0 capability-scoped profile, provide the supported inspection, diagnostics, compatibility guidance, and exact recipes qualified for that profile. Do not treat an absent Software 2.0-only service as a fault.
  6. Before any platform action, require its named capability to be true. If it is false, explain the profile restriction separately from relevant read-only guidance.
  7. If version_specific_guidance is false or retrieval abstains, say no applicable version-specific passage was found and avoid inventing a platform command.
  8. Answer from the combined profile and applicable passage. Label unknown evidence as unknown.

Safety requirements

  • Before any mutating action, display the exact command to be run and obtain explicit user approval immediately before executing it. Never batch approvals or carry one forward to a later action.
  • Never assume an nvidia-smi index is a CUDA ordinal.
  • Use GPU UUIDs for every proposed launch. When multiple GPUs are deliberately visible, place the GB300 UUID first.
  • Do not apply either legacy or Software 2.0 mixed-device behavior without checking the detected profile and observed ATS/HMM state.
  • Do not tell the user to install or operate Fabric Manager as a normal Station requirement.
  • When mixed_coherency_service is true, inspect mixed-coherency-gpu-select.service, its generated environment, and container exposure separately.
  • When dynamic_power_sloshing is true, inspect vsloshd. Always inspect observed caps and never propose an ad hoc power-cap change.
  • Never install or change the driver, kernel, OS, firmware, or packages.
  • Never display credential values.

Read references/platform.md when interpreting coherency, container ordering, power, memory placement, or qualification evidence. Read references/software-compatibility.md before applying version-specific guidance. Read references/sources.md when guidance may come from the Development Guide or bring-up guide. Read references/cli.md for command and JSON details.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.