Use when uploading images, labeling, organizing datasets, creating Roboflow projects (detection/segmentation/keypoint/classification), tags, splits, versions, or RoboQL search.
88
—
Does it follow best practices?
Impact
99%
1.45xAverage score across 3 eval scenarios
Advisory
Suggest reviewing before use
Dataset version generation with preprocessing and augmentation pipeline
Frozen snapshot understanding
33%
100%
Filter by Tag preprocessing
60%
100%
Tile preprocessing step
90%
100%
Filter Null step
60%
100%
Modify Classes step
60%
100%
Auto-Orient preprocessing
87%
100%
Augmentation train-only scope
100%
100%
3x augmentation multiplier
75%
100%
versions_generate or SDK call
100%
100%
Non-destructive class change
70%
100%
Train/valid/test split configuration
100%
100%
Preprocessing vs augmentation distinction
100%
100%
RoboQL query construction with boolean logic and specialized filters
split: filter usage
100%
100%
class: with comparison operator
100%
100%
NOT or minus inversion
100%
100%
min-annotations filter
100%
100%
min-width filter
62%
100%
like-image filter
0%
100%
filename wildcard
25%
100%
project: filter
100%
100%
Boolean AND combination
62%
100%
Semantic/free-text search
100%
100%
images_search MCP or SDK
100%
87%
Query syntax comments
100%
100%
Multi-project setup with CLI upload, tagging, and annotation group sharing
CLI for large dataset
50%
100%
CLI install step
100%
100%
Correct CLI syntax
25%
100%
Web UI for small dataset
87%
100%
Project type immutability
25%
100%
Annotation group sharing
70%
100%
Shared annotation side-effects
37%
100%
Tag-based image organization
100%
100%
Tag application method
87%
100%
Duplicate image handling
0%
100%
Image size limits
0%
100%
Annotation group at creation only
33%
100%
Shared image usage counting
0%
100%
bf3aefb
Table of Contents
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.