HuggingFace Agent Skills: Standardized AI Task Definitions for Coding Agents

HuggingFace Skills are standardized definitions for AI/ML tasks that work with major coding agent tools. Each skill packages instructions, scripts, and resources in a folder with a SKILL.md file containing YAML frontmatter (name and description) followed by guidance for the coding agent.

How Skills Work

Skills follow a standardized format and are compatible with multiple agent tools. The repository includes support for different agent systems:

• Claude Code uses the term "Skills" and requires plugin registration
• OpenAI Codex uses an AGENTS.md file for instructions
• Google Gemini uses 'extensions' defined in gemini-extension.json
• Cursor uses plugin manifests (.cursor-plugin/plugin.json and .mcp.json)

Installation Methods

Claude Code:
/plugin marketplace add huggingface/skills
/plugin install <skill-name>@huggingface/skills
Example: /plugin install hugging-face-cli@huggingface/skills

Codex:
Codex identifies skills via AGENTS.md. Verify with:
codex --ask-for-approval never "Summarize the current instructions."

Gemini CLI:
Install locally:
gemini extensions install . --consent
Or from GitHub:
gemini extensions install https://github.com/huggingface/skills.git --consent

Cursor:
Install from repository URL via Cursor plugin flow. For contributors, regenerate manifests with:
./scripts/publish.sh

Available Skills

• hugging-face-cli: Execute Hugging Face Hub operations using the hf CLI. Download models/datasets, upload files, manage repos, and run cloud compute jobs.
• hugging-face-datasets: Create and manage datasets on Hugging Face Hub. Supports initializing repos, defining configs/system prompts, streaming row updates, and SQL-based dataset querying/transformation.
• hugging-face-evaluation: Add and manage evaluation results in Hugging Face model cards. Supports extracting eval tables from README content, importing scores from Artificial Analysis API, and running custom evaluations with vLLM/lighteval.
• hugging-face-jobs: Run compute jobs on Hugging Face infrastructure. Execute Python scripts, manage scheduled jobs, and monitor job status.
• hugging-face-model-trainer: Train or fine-tune language models using TRL on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes hardware selection, cost estimation, Trackio monitoring, and Hub persistence.
• hugging-face-paper-publisher: Publish and manage research papers on Hugging Face Hub. Supports creating paper pages, linking papers to models/datasets, claiming authorship, and generating professional markdown-based research articles.
• hugging-face-tool-builder: Build reusable scripts for Hugging Face API operations.

If your agent doesn't support skills, you can use agents/AGENTS.md directly as a fallback. The repository is open for contributions of new skills.