tao-train-pose-classification

Pose classification using ST-GCN (Spatial Temporal Graph Convolutional Network). Classifies skeleton sequences into action categories from pose-keypoint data. Use when training, evaluating, exporting, or running inference for a TAO pose-classification model. Trigger phrases include "train pose classification", "skeleton action recognition", "ST-GCN", "keypoint sequence classifier".

Quality

91%

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

83%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 body is highly actionable with executable commands and concrete spec overrides, well-structured into clear sections that point to real reference files. It is slightly verbose with some duplicated guidance and relies on implicit rather than explicit validation checkpoints.

Suggestions

Remove the duplicated `--shm-size=8g` line in the DOCKER_COMMON array and consolidate the dataset_convert explanation so it appears once rather than across multiple sections.

Add explicit validation/retry checkpoints for batch and destructive operations (e.g. verify checkpoint existence and dataset path consistency before launching evaluate/export).

Either package the referenced schemas/<action>.schema.json and schemas/manifest.json files or clarify that they are generated at packaging time so the body's references resolve.

Dimension	Reasoning	Score
Conciseness	Mostly efficient with concrete commands and tables, but has padded/redundant elements (e.g. repeated `--shm-size=8g` line, near-duplicate dataset_convert explanations in multiple sections, and restated policy prose).	4 / 5
Actionability	Provides copy-paste-ready docker run commands, complete spec_override JSON blocks per action, and explicit checkpoint/data-path examples covering the common cases.	5 / 5
Workflow Clarity	Clear sequencing across dataset_convert→train→evaluate→export→inference with mandatory-data-source and checkpoint handoff guidance, but validation checkpoints are implicit (no explicit validate-then-retry loop) for these batch operations.	4 / 5
Progressive Disclosure	Well-organized sections referencing real bundle files (references/skill_info.yaml, references/spec_template_<action>.yaml) one level deep, though several reference paths (schemas/<action>.schema.json, schemas/manifest.json) are mentioned in the body but are absent from the provided bundle.	4 / 5
	Total	17 / 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 concrete, third-person, and explicitly pairs the 'what' with a 'Use when' trigger clause plus natural trigger phrases. It avoids fluff and is clearly distinguishable from sibling TAO model skills.

Dimension	Reasoning	Score
Specificity	Names the domain (pose classification via ST-GCN) and concrete actions — 'training, evaluating, exporting, or running inference' — with comprehensive coverage of the supported CLI actions.	5 / 5
Completeness	Explicitly answers 'what' (classifies skeleton sequences into action categories) and 'when' ('Use when training, evaluating, exporting, or running inference') with concrete trigger phrases.	5 / 5
Trigger Term Quality	Lists natural trigger phrases users would actually say: 'train pose classification', 'skeleton action recognition', 'ST-GCN', 'keypoint sequence classifier', covering synonyms and the model acronym.	5 / 5
Distinctiveness Conflict Risk	Niche is clearly defined (ST-GCN pose/keypoint classification) with distinct triggers unlikely to match other TAO model skills; minimal conflict risk.	5 / 5
	Total	20 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 2 missing	Warning

	Total	14 / 16 Passed

Repository: NVIDIA/skills
Path: skills/tao-train-pose-classification/SKILL.md
Commit: be3d1a4

Reviewed: 3 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.