Model Versioning Tracker Skill for Traceable Model Cards
Link model versions to runs, dataset identity, metrics, and model cards so a selection can be reviewed later.
中文Install in Claude Code
Use local MLflow and two synthetic model runs evaluated on the same test set and metric definition.
- Get the complete `plugins/ai-ml/model-versioning-tracker/skills/tracking-model-versions` directory from the pinned GitHub commit.
- Copy it into the target project's `.claude/skills/`. Check `.claude/skills/tracking-model-versions/SKILL.md`, start Claude Code at the project root, and call `/tracking-model-versions`.
Example task
Register two parcel model versions with different run IDs. Record MAE 3.2 and 2.8 on the same fixture and connect each version to its run.What to check
The model card links both runs and the same dataset reference. The report shows the 0.4 MAE difference and usage limits.
Original SKILL.md on GitHub · Updated
Frequently asked question
Can a lower MAE automatically replace a current model?
Verify agreed performance and usage conditions, including latency and representative inputs, before selecting the version.