ml-stack-hub
Manage Hugging Face or repository artifacts when a user asks to browse revisions, inspect cards, upload files, commit, open a PR, create a discussion, configure a webhook, or publish a model, dataset,
SKILL.mdskills/ml-stack-hub/SKILL.md
GitHubHub and repository lifecycle
Require artifact type and ID, exact revision, intended operation, license and metadata, visibility, files, provenance, destination, and explicit approval for every write or external notification.
Workflow
- Resolve model, dataset, Space, or repository identity to an exact URL and revision; distinguish mutable main branches from immutable commits/tags.
- Inspect cards, metadata, license, README, file manifest, and existing provenance before proposing changes.
- Keep traces and sensitive artifacts private by default; redact secrets and personal data.
- Treat uploads, commits, PRs, discussions, webhooks, publication, and visibility changes as separate approval-gated operations. If no matching runtime operation exists, return a bounded plan with exact API/CLI intent and expected response.
- Hand immutable URLs, hashes, metadata, and approval evidence to
ml-stack-deploymentorml-stack-audit.
Outputs and abstention
Never claim a write, PR, discussion, webhook, or publication occurred without an observed result. Abstain when revision, license, destination, or approval is missing. Next owner: ml-stack-deployment for serving release, ml-stack-tracking for private traces, or ml-stack-audit for final artifact reconciliation.
See Hub revisions, cards and metadata, repository and PR policy, and trace privacy.