nemo-curator

GPU-accelerated data curation for LLM training. Supports text/image/video/audio. Features fuzzy deduplication (16× faster), quality filtering (30+ heuristics), semantic deduplication, PII redaction, NSFW detection. Scales across GPUs with RAPIDS. Use for preparing high-quality training datasets, cleaning web data, or deduplicating large corpora.

Quality

86%

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

72%

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

The body is highly actionable and well-structured with clean progressive disclosure to real reference files. Its main weaknesses are verbosity (duplicated benchmark/cost content and redundant trailing sections) and missing validation checkpoints in the batch curation workflow.

Suggestions

Consolidate the duplicated performance figures: keep one 'Performance benchmarks' / 'Cost comparison' section and remove the restated 16x/120h numbers from 'Performance' and stage descriptions to tighten conciseness.

Add a validation checkpoint to the Common Crawl pipeline (e.g., 'Verify document count and inspect filtered sample before saving') so the batch/destructive workflow earns workflow_clarity 3.

Trim the low-signal 'Use cases' and 'Resources' sections (star counts and external links rarely earn their tokens in a skill body).

Dimension	Reasoning	Score
Conciseness	Mostly efficient code, but padded with duplicated benchmark/cost tables (Stage descriptions restate the 'Performance benchmarks' and 'Cost comparison' sections) plus low-value 'Use cases' and 'Resources' sections that could be trimmed.	2 / 3
Actionability	Imports, concrete parameters, and copy-paste-ready code blocks across text, image, video, and audio curation match the fully-executable anchor; not the level below because no pseudocode is used.	3 / 3
Workflow Clarity	The pipeline stages are clearly sequenced, but destructive/batch operations (dedup, filtering) lack any validation or verification checkpoint, which caps workflow_clarity at 2 per the rubric guidelines.	2 / 3
Progressive Disclosure	A concise overview in SKILL.md links to well-signaled, one-level-deep real bundle files (references/filtering.md, references/deduplication.md), matching the clear-overview anchor; not below because references are genuinely one level deep and easily navigable.	3 / 3
	Total	10 / 12 Passed

Description

100%

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 specific, trigger-rich, and explicitly pairs 'what' with 'when', hitting the top anchor on every dimension. It uses third person throughout and avoids vague fluff.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions—'fuzzy deduplication', 'quality filtering (30+ heuristics)', 'semantic deduplication', 'PII redaction', 'NSFW detection'—matching the score-3 anchor.	3 / 3
Completeness	Explicitly answers both what it does and when to use it via the 'Use for preparing high-quality training datasets, cleaning web data, or deduplicating large corpora' clause.	3 / 3
Trigger Term Quality	Natural user terms like 'cleaning web data', 'deduplicating large corpora', and 'preparing high-quality training datasets' give good coverage of how users would phrase the need.	3 / 3
Distinctiveness Conflict Risk	A clear niche (GPU-accelerated LLM training-data curation) with distinct triggers makes it unlikely to conflict with adjacent data-processing skills.	3 / 3
	Total	12 / 12 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: OpenLAIR/dr-claw
Path: skills/data-processing/nemo-curator/SKILL.md
Commit: 6da7f7c

Reviewed: about 7 hours 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.