HF Skill Pack

Hugging Face's skills repository is the most ML/agent-oriented entry in the collection — skills tailored for working with Hugging Face Spaces, dataset operations, model evaluation harnesses, and agent chaining patterns. Pairs with the existing Chaining HF Spaces for Agent Workflows concept page on this site.

When to use

• You're building agents that interact with HF Spaces or HF Inference API
• You need a skills baseline for ML evaluation workflows
• You want to chain Spaces together as agent tools (the HF-specific version of an MCP-style composition)

When not to use

• You're not in the HF ecosystem — the skills are cross-compatible but the value is HF-specific
• You need general-purpose coding skills — see Superpowers or Anthropic Official instead

How it works

Standard SKILL.md + progressive-disclosure pattern. The skills are designed to compose with the existing HF tooling (the huggingface_hub Python library, the Spaces API) rather than replace it. Several skills in the pack are specifically tuned for the agent-chaining pattern this site documents.

Install

npx skills add huggingface/skills

Full SKILL.md

Full source at the canonical repo: github.com/huggingface/skills. Embedded copy omitted at launch; the canonical source is authoritative.

Verify it yourself

Quick static check — HF skills may legitimately call the HF API; confirm any network calls in scripts are going to huggingface.co domains and not to third-party endpoints. Bounded allowed-tools declarations are a positive signal.

Two-model review — paste each SKILL.md you intend to use into two different models with the Vetting Guide prompt.

Review transparency

Related

• Chaining HF Spaces for Agent Workflows
• HF Spaces Multi-Agent Chain — the blueprint
• Skills Collection Overview
• Vetting Guide