nemotron-3-ultra-text2sql-lora

Run the Nemotron-3 Ultra Text2SQL LoRA fine-tuning tutorial (NeMo Megatron-Bridge) end-to-end for the user on their SLURM cluster: data prep, distributed checkpoint conversion, and packed LoRA fine-tuning of the 550B hybrid Mamba-Transformer MoE, ending at a saved adapter. Use when the user wants to run this cookbook, fine-tune Nemotron-3 Ultra with LoRA, or adapt the notebook to their own cluster.

Quality

85%

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

Quality

Content

85%

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

A tightly written, well-sequenced runbook that assumes competence and provides strong validation checkpoints across all three SLURM phases. Its only real gap is actionability: the executable commands live inside notebook cells and placeholders rather than as copy-paste-ready shell in the body itself.

Suggestions

Add a short, concrete example of the key sbatch/squeue invocations or the minimal config.env key set inline, so an agent can act without first opening the notebook cells.

Since the notebook and slurm scripts are referenced but not present in the bundle, note explicitly that those files ship alongside this SKILL.md so the references are verifiable as progressive disclosure.

Dimension	Reasoning	Score
Conciseness	The body is lean and assumes Claude's competence — it states cluster/scale facts and the per-step rhythm without padding or explaining concepts Claude already knows; every section earns its place.	3 / 3
Actionability	It gives concrete paths, env keys, and SLURM commands (`sbatch`, `squeue`, `sacct`, `config.env` keys, output paths) but the actual run instructions are mediated by notebook cells and placeholders rather than copy-paste-ready commands, leaving the literal execution partly abstract.	2 / 3
Workflow Clarity	The three-step sequence is clearly ordered with an explicit launch/check/sanity loop per step, idempotency noted, expected outputs defined, and friction/misread cases flagged — a full feedback-checkpoint workflow for risky multi-node jobs.	3 / 3
Progressive Disclosure	The body is a single-level overview pointing to in-directory artifacts (`mbridge_lora_cookbook.ipynb`, `config.env`, `slurm/*.sbatch`) with no nested reference chains; bundle references are appropriately one level deep and clearly signaled, even though those files are not bundled here.	3 / 3
	Total	11 / 12 Passed

Description

85%

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

A strong, specific description with an explicit Use-when clause and a clear three-phase end state. Its main weakness is reliance on the internal term 'cookbook' in the trigger phrasing, which slightly limits natural trigger coverage.

Suggestions

Broaden the trigger clause with user-natural phrasing such as 'fine-tune a Nemotron model for SQL generation' or 'LoRA fine-tuning on a SLURM cluster' rather than leaning on the internal 'cookbook' label.

Consider mentioning the cluster/SLURM context earlier in the description so the multi-node SLURM requirement is visible at trigger-evaluation time.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'data prep, distributed checkpoint conversion, and packed LoRA fine-tuning ... ending at a saved adapter' — each tied to a specific deliverable rather than vague phrasing.	3 / 3
Completeness	Explicitly answers both 'what' (run the tutorial end-to-end through three named phases ending at a saved adapter) and 'when' via an explicit 'Use when...' clause, so both halves are clearly present.	3 / 3
Trigger Term Quality	'Use when the user wants to run this cookbook, fine-tune Nemotron-3 Ultra with LoRA, or adapt the notebook' covers several natural triggers but is narrowly framed around an internal tutorial name ('cookbook') that a user is unlikely to say verbatim, missing broader variations.	2 / 3
Distinctiveness Conflict Risk	The skill targets a very specific niche (Nemotron-3 Ultra Text2SQL LoRA on Megatron-Bridge via SLURM) with distinct triggers, making it unlikely to fire for unrelated skills.	3 / 3
	Total	11 / 12 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: usage-cookbook/Nemotron-3-Ultra/lora-text2sql/nemo-megatron-bridge/SKILL.md
Commit: 5277655

Reviewed: 2 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.