dataset-annotation

AI-assisted dataset annotation with COCO export — bbox, SAM2, DINOv3 methods

Quality

61%

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/annotation/dataset-annotation/SKILL.md

Quality

Content

72%

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

A tight, well-structured skill body that excels at conciseness and organization, but its annotation workflow is described rather than made executable and lacks the validation checkpoints expected for a batch dataset-export operation.

Suggestions

Add an explicit validation step to the Annotation Loop (e.g. 'Verify COCO schema with scripts/validate.py before saving' or 'Confirm annotation count matches detections before export') to earn a workflow-clarity top score.

Show a concrete invocation of scripts/annotate.py or a minimal executable snippet for one method (bbox or SAM2) so the core action is copy-paste ready rather than described.

Link the body to the existing scripts/annotate.py bundle (e.g. a '## Implementation' pointer) so readers can navigate from the overview to the executable script.

Dimension	Reasoning	Score
Conciseness	The ~45-line body is lean and efficient: it lists capabilities tersely and gives compact JSONL protocol examples without explaining what COCO, SAM2, or DINOv3 are, so every token earns its place.	3 / 3
Actionability	The JSONL protocol spec and the venv/pip setup snippet are concrete and copy-paste ready, but the core annotation workflow is described ('draw bounding boxes', 'click to segment') rather than given as executable code, and the body never shows how to invoke scripts/annotate.py.	2 / 3
Workflow Clarity	The five-step 'Annotation Loop' is clearly sequenced, but it contains no validation or verification checkpoints for what is a batch export operation, which caps workflow clarity at 2 per the rubric's destructive/batch guideline.	2 / 3
Progressive Disclosure	Under 50 lines with well-organized, clearly headed sections (What You Get, Annotation Loop, Protocol, Setup) and a self-contained body with no nested references, satisfying the simple-skill allowance for a top score.	3 / 3
	Total	10 / 12 Passed

Description

50%

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 concise, third-person description that names a specific niche and concrete methods, but it lacks any 'Use when' trigger guidance and leans on jargon (SAM2, DINOv3) over natural user phrasing, leaving every dimension at the middle anchor.

Suggestions

Add an explicit 'Use when...' clause listing natural triggers, e.g. 'Use when annotating images for detection datasets, labeling bounding boxes, or exporting COCO/YOLO datasets for model training.'

Soften the jargon-heavy method list with plain terms users would actually say ('bounding boxes', 'segmentation masks', 'visual grounding') to improve trigger-term coverage.

Expand the action verbs beyond 'annotation' and 'export' to name concrete operations (draw boxes, segment objects, track across frames, export datasets) for stronger specificity.

Dimension	Reasoning	Score
Specificity	Names the annotation domain and COCO export action plus three method names ('bbox, SAM2, DINOv3 methods'), but the actions themselves are limited to 'annotation' and 'export' rather than a comprehensive set of concrete verbs like the anchor-3 example.	2 / 3
Completeness	It clearly states what the skill does ('AI-assisted dataset annotation with COCO export') but provides no 'Use when...' clause or equivalent explicit trigger guidance, which caps completeness at 2 per the rubric guidelines.	2 / 3
Trigger Term Quality	'dataset annotation' and 'COCO' are natural practitioner terms, but common variations a user would say (image labeling, bounding boxes, segmentation, object detection datasets) are absent and the em-dash method list reads as a technical spec.	2 / 3
Distinctiveness Conflict Risk	The dataset-annotation-with-COCO niche is fairly specific and unlikely to conflict with unrelated skills, but without explicit triggers it could still overlap with generic image-labeling or computer-vision skills.	2 / 3
	Total	8 / 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: SharpAI/DeepCamera
Path: skills/annotation/dataset-annotation/SKILL.md
Commit: 2264fcb

Reviewed: 1 day 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.