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.

Quality

77%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./skills/nemotron-super3/SKILL.md

Quality

Content

67%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is a well-structured, actionable reference-desk skill with a clear Locate→Retrieve→Cite workflow, concrete file routing, and useful calibration examples. Its main weakness is conciseness: the Source hierarchy section and calibration examples duplicate guidance already given in the routing table and answering rules.

Suggestions

Collapse or remove the "Source hierarchy" section — its file ordering and per-file descriptions repeat the "Locate" routing table and "Retrieve" pattern; a one-line pointer would suffice.

Trim the four calibration examples to one or two, or move them into a referenced file; they largely restate rules already in "Answering rules" and "Known caveats".

Add an explicit Cite-stage validation step (e.g. "confirm the cited file path actually contains the claim before answering") to push workflow clarity from a checkpoint toward a true verify loop.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient skill-specific routing guidance (it does not explain concepts Claude already knows like what MoE or RL is), but the "Source hierarchy" section largely duplicates the "Locate" routing table and "Retrieve" pattern, and the four calibration examples restate rules from the "Answering rules" section, so it could be tightened past anchor-3 toward anchor-4.	3 / 5
Actionability	Provides highly concrete, specific guidance — a question-to-file routing table, exact file paths (e.g. "paper/rl/rlvr.md", "recipes/stage2_rl_swe2.md"), a citation format, and four worked Q&A examples covering common cases; it stops short of anchor 5 only because, as an instruction-only skill, it offers no executable commands, which the anchor-5 example centers on.	4 / 5
Workflow Clarity	Sequences the workflow explicitly ("Locate → Retrieve → Cite. Always work in this order") with numbered sub-steps and a Cite checkpoint that catches source disagreements, plus Do/Do-not and caveats checklists; it lacks an explicit output-validation step (e.g. verifying the cited file actually supports the claim), keeping it below anchor 5.	4 / 5
Progressive Disclosure	SKILL.md is a well-signaled overview with one-level-deep references (routing table maps questions to paper/.md and recipes/.md files) and easy navigation, matching anchor-4; it does not reach anchor 5 because the referenced bundle files are not actually present to verify, and the long inline routing table overlaps content that could live in INDEX.md/quick-reference.md.	4 / 5
	Total	15 / 20 Passed

Description

87%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is strong: it clearly states both what the skill does and when to use it, with a concrete trigger and a disambiguating negative trigger that limits conflict with sibling skills. The only weakness is slightly topical rather than action-oriented verb framing and a few missing natural synonym triggers.

Dimension	Reasoning	Score
Specificity	Lists several specific coverage areas — "architecture, training data, recipes (pretrain/SFT/RL/eval/quantization), and deployment notes" — with comprehensive topic coverage, but the verbs stay topical ("reference desk for...") rather than enumerating concrete actions, so it sits just below the anchor-5 example.	4 / 5
Completeness	Explicitly answers both what ("Reference desk for NVIDIA Nemotron 3 Super — architecture, training data, recipes ... and deployment notes") and when ("Use when the user asks facts about Super3 rather than building a pipeline") with a concrete trigger phrase, matching the anchor-5 example.	5 / 5
Trigger Term Quality	Includes a natural trigger ("Use when the user asks facts about Super3") plus a useful negative trigger ("rather than building a pipeline"), giving good keyword coverage; it misses common synonyms/variants such as "Nemotron", "benchmark", or "quantization" as standalone triggers, so it does not reach anchor 5.	4 / 5
Distinctiveness Conflict Risk	Has a clear niche (Nemotron 3 Super facts) and an explicit disambiguating negative trigger that hands pipeline-building off to another skill, giving minimal conflict risk per the anchor-5 example.	5 / 5
	Total	18 / 20 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: NVIDIA-NeMo/Nemotron
Path: skills/nemotron-super3/SKILL.md
Commit: 1a4c3fa

Reviewed: 4 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.