SKILLEMALL.ai
GitHub ↗ RU Sign in

BC nemotron-super3

Reference desk for NVIDIA Nemotron 3 Super — architecture, training data, recipes (pretrain/SFT/RL/eval/quantization), and deployment notes. Use when the user asks facts about Super3 rather than building a pipeline.

The skillemall take

Reference for NVIDIA Nemotron 3 Super: architecture, training recipes (pretrain, SFT, RL, eval, quantization), and deployment. Contains 26 files with 2320 tokens of documentation.

Grade B, quality score 84, safety 96 indicate solid shape. Process score sits at 63, below average but no critical issues found. Four medium/low findings don't affect usability for a reference. Runs on all major platforms from Claude to DeepSeek.

Install if you're working with Nemotron training pipelines and need NVIDIA's official recipes. Skip if you need hand-holding.

NVIDIA-NeMo/Nemotron Agent Skills author: NVIDIA-NeMo 26 files body ≈ 2 320 tokens Open the sourcegithub.com↗ analyzed 5 d ago

Reference desk for NVIDIA Nemotron 3 Super — architecture, training data, recipes (pretrain/SFT/RL/eval/quantization), and deployment notes.

As a process C 63/100 · Has gaps — weak spots: result and completion, inputs and preconditions, running it twice

ReferenceInfrastructuretype and topics are labelled automatically from the skill text
Runs in: Claude / Claude Code ChatGPT / Codex Gemini CLI Cursor GitHub Copilot / VS Code Qwen Code Kimi Code CLI Z.ai (GLM) DeepSeek (Deep Code) Mistral Vibe LangChain (deepagents) OpenClaw Goose OpenCode
JSON
Technical rating
B
91/100
safety, quality, tests
Safety 60%
96
Quality 40%
84
Run on models
none yet
Process rating
C
63/100
Has gaps
Result and completion w 14
0
Inputs and preconditions w 11
0
Progress reporting w 2
0
the three weakest of ten parameters · all ten

How to improve

    For the model run — optional
    • Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
    • A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.

    Guard findings · 4

    ✓ No critical or high findings

    Medium and low: 4
    • low Secrets in code secret-high-entropy-token context/quick-reference.md:293
      High-entropy token-like string (may be an id, hash or a credential)
      - init: **Qwen…507**
    • low Secrets in code secret-high-entropy-token model-card.md:11
      High-entropy token-like string (may be an id, hash or a credential)
      | Model | NVID…12B |
    • low Secrets in code secret-high-entropy-token model-card.md:43
      High-entropy token-like string (may be an id, hash or a credential)
      - Qwen…603 (used as the GenRM in RLHF)
    • low Secrets in code secret-high-entropy-token paper/rl/rlhf.md:54
      High-entropy token-like string (may be an id, hash or a credential)
      The paper and recipe docs say it is initialized from **Qwen…507**.

    Files scanned: 26. Evidence is masked. Grey chips explain why severity was lowered.

    Against the Agent Skills spec

    ✓ No remarks against the Agent Skills spec

    Process rating: all ten parameters 63/100

    • 0Result and completion. Does not say what the result is
    • 0Inputs and preconditions. Does not say what the process needs to start
    • 0Progress reporting. Says nothing while it works
    • 30Running it twice. 2 mutating operations with no state check
    • 65Failures and branches. 3 branches
    • 70When it triggers. States when to use, but not when not to
    • 100Tools and files. No external tools needed
    • 100Steps. 90 steps
    • 100Consistency. Name and required fields are in place
    • 100Execution cost. Instruction body is 2320 tokens

    Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.

    Quality signals

    • +5Description has no quoted example phrases that should trigger the skill
    • +4Description does not say when NOT to use the skill (false activations)
    • +3Output format is not stated: the model decides each time
    • +1No license
    • +2Single-language instructions
    • +3Description length 215: enough signal without eating the budget
    • +4Structure: 21 headings
    • +3Step-by-step instructions: 90 items
    • +4Has examples (0 code blocks)

    Quality base 70; lint remarks subtract, signals add up to 100. Result: 84.