cvat-push
Skill8 starsChanged 8 months agoMITView on GitHub
Push uncertain or misclassified images to CVAT for human annotation review after training analysis.
- Reads credentials
In this repository15 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- active-learningSkill
- analyze-resultsSkill
- annotateSkill
- auto-experimentSkill
- benchmarkSkill
- compare-modelsSkill
- cvat-deploySkill
- cvat-pullSkill
- explain-resultsSkill
- monitor-trainingSkill
- review-annotationsSkill
- review-datasetSkill
- setup-projectSkill
- training-loopSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
After training analysis identifies uncertain images, push them to CVAT for human annotation or correction.
- Distinctive: line 19
Read the analysis output:
- Distinctive: line 69
NEVER push more than 200 images without asking the user first
- Distinctive: line 70
ALWAYS include pre-annotations when available (reduces annotation time)
- Distinctive: line 71
Keep a local copy of all images — CVAT is not a backup
- Distinctive: line 72
Write the push report before telling the user it's done
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Push to CVAT for Review
After training analysis identifies uncertain images, push them to CVAT for human annotation or correction.
Pre-Flight Checklist
yolo-analyzehas been run (check foruncertain_images.txtor analysis output)- CVAT configured in
yolo-project.yaml CVAT_ACCESS_TOKENenv var is set
Workflow
1. Gather Uncertain Images
Read the analysis output:
cat reports/uncertain_images.txtOr check experiments/analysis.md for the list of uncertain predictions.
2. Categorize by Priority
Organize images into review batches:
- False negatives (highest priority): Images where the model missed detections but labels exist
- Low confidence (medium priority): Detections with confidence between low_conf and high_conf thresholds
- No detection, no label (low priority): Images where neither model nor labels found anything
3. Push to CVAT
# From analysis file (auto-batches into tasks of 50)
yolo-cvat push --from-analysis reports/uncertain_images.txt
# Or manually specify images
yolo-cvat push --images <path_to_images> --labels <path_to_labels> --task-name "Review: False Negatives"4. Generate Report
Write experiments/cvat_push_report.md:
## CVAT Upload Report
- Date: YYYY-MM-DD
- Tasks created: N
- Total images: N
- False negatives: N (Task ID: X)
- Uncertain: N (Task ID: Y)
- CVAT URL: http://localhost:8080/tasks/<id>Decision Tree
More than 100 images?
├── Yes → Split into batches of 50, create multiple tasks
└── No → Single task
Has existing YOLO labels for these images?
├── Yes → Upload as pre-annotations (saves annotator time)
└── No → Push images only for fresh annotation
False negatives found?
├── Yes → Create separate high-priority task labeled "URGENT: False Negatives"
└── No → Single "Review: Uncertain" taskGuardrails
- NEVER push more than 200 images without asking the user first
- ALWAYS include pre-annotations when available (reduces annotation time)
- Keep a local copy of all images — CVAT is not a backup
- Write the push report before telling the user it's done
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

