writing-model-cards

To write a production-ready model card, document intended use, out-of-scope uses, provenance, subgroup metrics, and traceability. The skill drafts these sections with 95% confidence intervals, concrete harms, mitigations, and an AIBOM supply-chain addendum for defensible compliance.

An AIBOM is an AI Bill of Materials addendum capturing supply-chain traceability for a model. It documents training and evaluation provenance, model details, and subgroup factors, ensuring auditor-ready compliance for deployment-bound, open-weights, and high-stakes review scenarios.

Document subgroup metrics and confidence intervals by capturing evaluation provenance and subgroup factors within the model card. The skill requires 95% confidence intervals for subgroup metrics, ensuring defensible documentation for compliance review and auditor readiness.

Yes, this model card template works for open-weights releases and compliance reviews. It captures intended use, explicit out-of-scope uses, ethical considerations, and limitations, making it suitable for high-stakes models like fraud detectors or medical classifiers needing auditor-ready documentation.

The best way to document out-of-scope uses and ethical considerations is using a Mitchell-style model card structure. The skill drafts explicit out-of-scope uses, concrete harms, and mitigations, turning a vague checklist into complete, defensible documentation for deployment.

You need an AIBOM supply-chain addendum to ensure traceability and compliance for deployment-bound models. It captures training and evaluation provenance alongside model details, providing auditor-ready documentation required for high-stakes AI governance and compliance review scenarios.