How to use the HF Paper Publisher agent skill

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file with instructions, plus optional scripts, templates and reference files that an AI agent can load when a task matches the skill's description. This skill is described as publishing and managing research papers on the Hugging Face Hub.

In practice, the agent can index an arXiv paper as a Hugging Face Paper Page, add paper references to a model or dataset README so the Hub tags the repository with the paper, start an authorship claim, toggle which papers show on your profile, and generate starter markdown research articles from four templates that can be converted to HTML.

• Index arXiv papers as Hugging Face Paper Pages
• Link papers to model or dataset repositories
• Start an authorship claim (admins review it)
• List your papers and show or hide them on your profile
• Create markdown research articles from standard, modern, arxiv or ml-report templates
• Convert a markdown paper to HTML

Install the HF Paper Publisher skill

Download the ZIP for the skill from this page and unzip it. For a personal install, place the unzipped folder inside ~/.claude/skills/. For a single project, place it inside that project's .claude/skills/ folder.

You can also upload the skill in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the skill there.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills

Set up what the skill needs before running it

The skill runs its script with uv, so you need Python 3 and uv installed. The script resolves its own Python packages from its header, so you do not need to create an environment by hand.

Set the HF_TOKEN environment variable to a Hugging Face token with write access. Most actions need this token, so anyone without one will not get far. The skill also needs internet access to huggingface.co and arXiv.

• Python 3 with uv installed
• HF_TOKEN environment variable containing a Hugging Face token with write access
• Internet access to huggingface.co and arXiv
• Packages declared in the script header: huggingface_hub, pyyaml, requests, python-dotenv (SKILL.md also lists markdown)

How to trigger the skill with natural requests

You do not need to type the script commands yourself. Ask the agent in plain language and mention the paper, repository or outcome you want. Requests that name an arXiv ID, a Hugging Face repository, or a paper template are the clearest. The skill's own instructions say to run scripts from the skill's directory, so the agent should handle the working directory for you.

Good requests name the action and the object, for example indexing a specific arXiv paper, or linking that paper to a repository you own.

• "Index arXiv paper 2301.12345 on Hugging Face Papers."
• "Link arXiv 2301.12345 to my model username/my-awesome-model."
• "Link arXiv papers 2301.12345 and 2302.67890 to my dataset username/my-dataset."
• "Start an authorship claim for arXiv 2301.12345 using my institution email."
• "Show arXiv 2301.12345 on my profile."
• "Create a modern template research paper titled Fine-Tuning Large Language Models with LoRA."

Examples of common workflows

The included example document shows a typical sequence: check whether a paper already exists, index it, then fetch its information. The check command returns whether the paper exists along with its Hugging Face and arXiv URLs.

To link a paper, the script adds the arXiv reference to the repository README. The Hub then extracts the arXiv ID and adds an arxiv:<PAPER_ID> tag, which lets people click through to the Paper Page.

Create and convert a starter research article

The create command generates a markdown paper from one of four templates: standard, modern, arxiv or ml-report. The modern template is described as a clean, web-friendly layout with a dynamic table of contents and LaTeX math rendering. The convert command turns the markdown into HTML.

Treat the output as a starting layout. The template contains placeholder text such as bracketed prompts, so you still need to write the content.

• standard: traditional scientific paper structure
• modern: clean, web-friendly format inspired by Distill
• arxiv: arXiv-style formatting
• ml-report: machine learning experiment report

Safety notes: scripts and network access

This skill runs a bundled Python script and calls the Hugging Face API with your write-access token, so it can change repositories in your account. Only install it if you are comfortable with that. Review the script and the requests the agent plans to make before approving them, especially link and toggle commands that change repositories or profile visibility.

Claims are not confirmed by the skill. It only starts the process, and Hugging Face admins review it. Publishing a finished peer-reviewed paper to arXiv or a journal is outside the skill's scope.

• Keep HF_TOKEN in your environment and avoid pasting it into chats
• Check repository names before approving link commands
• The skill cannot confirm authorship; admins review claims

Tips for better results

Give the exact arXiv ID and the exact repository ID, using the form username/repo-name. Check the paper's Paper Page after indexing to confirm it looks right. If you want changes to a repository to go through a pull request, the link command documents a --create-pr option.

• Use the exact repository ID in the form username/repo-name
• Use --citation with a citation file if you want a custom citation added
• Link several papers at once with a comma-separated --arxiv-ids list
• Use the Settings → Papers section to toggle "Show on profile" in the web interface