Open-source ML research platform to train, fine-tune, evaluate and run models locally or on Slurm and cloud GPU clusters.
huggingface/skills
A collection of Agent Skills that give AI coding agents instructions and scripts for Hugging Face Hub tasks like training, evaluation, datasets and Spaces.
- Stars
- 11,288
- Forks
- 786
- Open issues
- 70
- Last commit
- 8 days ago
What is it?
Hugging Face Skills is a set of self-contained skill folders that package instructions, scripts and resources for AI coding agents working on machine learning tasks. Each skill is a directory with a SKILL.md file that the agent loads when the task matches, covering areas such as the hf command-line tool, model training with TRL, dataset exploration through the Dataset Viewer API, running evaluations, building Gradio demos and Spaces, memory estimation for model weights, and AWS SageMaker deployment.
The repository follows the Agent Skills format and works with Claude Code, Codex, Gemini CLI and Cursor. The hf-cli skill is the recommended starting point for core Hub operations such as searching models, managing datasets and buckets, and launching jobs. Additional workflow skills can be installed on demand with the hf skills add command.
Key features
- 01Ships an hf-cli skill that teaches agents the hf commands for Hub operations.
- 02Covers model training and fine-tuning with TRL and Unsloth on Hugging Face Jobs.
- 03Provides dataset querying, pagination and filtering through the Dataset Viewer API.
- 04Includes skills for building and deploying Gradio apps and Hugging Face Spaces.
- 05Offers an hf-mem skill to estimate memory needed to load Safetensors or GGUF weights.
- 06Includes AWS SageMaker deployment skills for planning, IAM setup and endpoints.
Use cases
- Letting a coding agent manage models, datasets and Spaces on the Hub from the terminal.
- Fine-tuning a language or vision model with guidance from a packaged training skill.
- Estimating GPU memory before loading a large model for inference.
- Deploying a Hub model to Amazon SageMaker with production defaults.
Quick start
- 1
Register the repository as a Claude Code plugin marketplace
/plugin marketplace add huggingface/skills - 2
Install the hf-cli skill
/plugin install hf-cli@huggingface/skills - 3
Install any additional Hugging Face skill with the hf CLI
hf skills add <skill-name>
Things to know
- Install instructions differ per client, so Codex, Gemini CLI and Cursor need manual setup steps.
In collections
Topics
FAQ
- What is skills?
- A collection of Agent Skills that give AI coding agents instructions and scripts for Hugging Face Hub tasks like training, evaluation, datasets and Spaces. Hugging Face Skills is a set of self-contained skill folders that package instructions, scripts and resources for AI coding agents working on machine learning tasks. Each skill is a directory with a SKILL.md file that the agent loads when the task matches, covering areas such as the hf command-line tool, model training with TRL, dataset exploration through the Dataset Viewer API, running evaluations, building Gradio demos and Spaces, memory estimation for model weights, and AWS SageMaker deployment.
- Is skills free to use? What license does it have?
- skills is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is skills still maintained?
- Actively maintained. The last commit to skills was on Oct 30, 2026.
- What can I use skills for?
- Letting a coding agent manage models, datasets and Spaces on the Hub from the terminal. Fine-tuning a language or vision model with guidance from a packaged training skill. Estimating GPU memory before loading a large model for inference. Deploying a Hub model to Amazon SageMaker with production defaults.
- How do I install skills?
- Register the repository as a Claude Code plugin marketplace: /plugin marketplace add huggingface/skills Install the hf-cli skill: /plugin install hf-cli@huggingface/skills Install any additional Hugging Face skill with the hf CLI: hf skills add <skill-name>
- Where can I get skills?
- The source code and installation instructions are on GitHub: https://github.com/huggingface/skills
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A toolkit for merging pre-trained large language models by combining their weights with many merge algorithms, run on CPU or GPU.
PyTorch DTensor-native training library for fine-tuning and pretraining LLMs, VLMs and diffusion models directly from Hugging Face checkpoints.
A config-driven CLI for running and comparing LLM fine-tuning experiments across prompts, open-source models, and LoRA settings, with built-in output tests.
SliceGPT is a post-training compression method that shrinks transformer models like LLMs by slicing rows and columns from weight matrices.
A hands-on Hugging Face course on aligning small language and vision-language models with minimal local hardware.