Draft Markdown model cards for ONNX tiny-AI models with provenance validation.
npx skills add https://github.com/VMAFx/vmafx --skill dev-llm-modelcard
Please help me install this Agent Skill. Skill: dev-llm-modelcard Source: https://github.com/VMAFx/vmafx/tree/main/.claude/skills/dev-llm-modelcard Command: npx skills add https://github.com/VMAFx/vmafx --skill dev-llm-modelcard
This Skill turns a shipped ONNX tiny-AI model into a reliable Markdown model card by gathering provenance, contract details, and measured facts before any prose is generated. It helps prevent unsupported claims and keeps documentation aligned with the actual artifact.
Use the dev-llm-modelcard skill to inspect the ONNX file, verify its provenance, and draft a Markdown model card for review.
High-intent search queries and answers about installing and using this skill.
To generate an ONNX model card, you need a validated ONNX path and fact blocks. The skill inspects the model, verifies replay evidence from provenance sidecars, and drafts a Markdown card for review.
Model provenance auditing verifies replay evidence from run sidecars to ensure documentation matches the actual artifact. It flags missing run provenance as a documentation gap before generating any prose.
The model card generator prevents inventing unsupported fields by using fact blocks. It collects graph details, sidecar provenance, and measured metrics first, ensuring the LLM only renders validated data.
You can review the facts before publishing because the workflow shows a facts-only preview first. After checking graph details and provenance, you can save, copy, or regenerate the draft.
If your ONNX model is missing run provenance sidecars, the documentation workflow flags the absent replay evidence as a gap. It still drafts the card using available graph details but notes the missing validation.
VMAFxAdd replay-manifest sidecars to AI-generated artifacts with run provenance.
yocxy2Audit markdown files and update PR metadata for software releases.
rocklambrosDraft production-ready model cards with intended use, metrics, and AIBOM details.
jassicsReview MLOps pipelines for data trust, access control, secrets, and model provenance.
