nemotron-customize

Plan, configure, and chain repo-native Nemotron customization steps into single-step or multi-step pipelines: curation, translation, SFT/PEFT (AutoModel or Megatron-Bridge), pretraining/CPT, RL alignment (DPO/RLVR/GRPO/RLHF), BYOB/MCQ benchmarks, checkpoint conversion, ModelOpt optimization, env profiles, and evaluation of trained checkpoints or existing/hosted endpoints. Use when a request names a Nemotron step or workflow, or asks to clean, translate, train, fine-tune, align, convert, optimize, evaluate, or compose these into a pipeline. Do NOT use for frontend/dashboard/visualization work, generic ML advice, billing/access, or non-Nemotron coding tasks.

Quality

92%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

85%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 routing-and-discipline skill: clear sequenced workflows with validation gates, strong progressive disclosure pointing to real bundle references, and concrete command shapes. Its main weakness is moderate redundancy across the Boundaries/Customization Surface/Operational Nuances sections that restate earlier rules.

Suggestions

Consolidate the overlapping rules in 'Boundaries', 'Customization Surface', and 'Operational Nuances' with 'Before You Begin' and 'Core Rule' to remove restated guidance about not editing checked-in configs and not inventing steps.

Trim the repeated 'never edit default.yaml/step.toml/runners; add a new config beside them' instruction, which currently appears in at least three sections.

Consider moving the Configuration Alignment bullet list into references/PATTERNS.md so the SKILL.md body stays a routing surface rather than an inline constraints catalog.

Dimension	Reasoning	Score
Conciseness	The body is dense and largely avoids explaining concepts Claude already knows, but sections like 'Boundaries', 'Customization Surface', and 'Operational Nuances' restate rules already covered in 'Before You Begin' and 'Core Rule', creating some redundancy that could be trimmed.	4 / 5
Actionability	It gives concrete, executable guidance (e.g. 'uv run nemotron steps show <step_id>', 'uv run nemotron steps run peft/automodel -c <config> --dry-run ...') and worked examples, though most commands stay parameterized rather than copy-paste ready, which is justified by the skill's requirement for user-provided concrete values.	4 / 5
Workflow Clarity	Multi-step flows are explicitly sequenced with checkpoints: the Orient->Plan->Act->Verify pipeline requires approval before writing, Single-Step Command Flow lists numbered steps with live verification, and destructive/remote operations require confirmation plus a 'Blocked' handoff when inputs are missing — strong validation feedback loops.	5 / 5
Progressive Disclosure	The body is a lean overview that signals one-level-deep references via a clear Reference Map table (CATALOG, ARTIFACTS, COMMANDS, HARDWARE, PATTERNS, WORKFLOW, context/index.toml, act/PROJECT, act/STAGE), all of which exist as real bundle files; detailed content is appropriately split into those references with explicit navigation.	5 / 5
	Total	18 / 20 Passed

Description

100%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 exemplary: it lists concrete capabilities, provides explicit 'Use when...' trigger guidance with natural terms, includes negative-scope carve-outs, and stays in third-person voice. It is dense but every clause earns its place for routing decisions.

Dimension	Reasoning	Score
Specificity	The description enumerates many concrete actions ('curation, translation, SFT/PEFT, pretraining/CPT, RL alignment (DPO/RLVR/GRPO/RLHF), BYOB/MCQ benchmarks, checkpoint conversion, ModelOpt optimization, env profiles, and evaluation'), giving comprehensive, specific coverage rather than vague language.	5 / 5
Completeness	It explicitly answers 'what' (plan, configure, and chain Nemotron customization steps) and 'when' with a concrete 'Use when...' clause listing triggering request shapes, plus an explicit 'Do NOT use for...' exclusion list.	5 / 5
Trigger Term Quality	It includes natural user verbs and nouns users would actually say ('clean, translate, train, fine-tune, align, convert, optimize, evaluate'), plus step-specific terms like 'benchmark' and 'smoke test', covering synonyms and common phrasings.	5 / 5
Distinctiveness Conflict Risk	The Nemotron-repo-native scope and explicit 'Do NOT use for frontend/dashboard/generic ML advice/billing/non-Nemotron coding' carve-out give it a clear niche with minimal conflict risk against other skills.	5 / 5
	Total	20 / 20 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_field	'metadata' should map string keys to string values	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 4 deeper-than-1-level	Warning

	Total	13 / 16 Passed

Repository: NVIDIA-NeMo/Nemotron
Path: skills/nemotron-customize/SKILL.md
Commit: 907d408

Reviewed: 6 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.