Look up and read Hugging Face paper pages in markdown, and use the papers API for structured metadata such as authors, linked models/datasets/spaces, Github repo and project page. Use when the user shares a Hugging Face paper page URL, an arXiv URL or ID, or asks to summarize, explain, or analyze an AI research paper.
huggingface-papers is a skill for looking up and reading AI research papers through Hugging Face Paper pages (hf.co/papers), a platform layered on top of arXiv. It lets an agent fetch a paper's full content as markdown or retrieve structured metadata — authors, linked model checkpoints, datasets and Spaces, GitHub repo, and project page — from a paper page URL, an arXiv URL or ID. The problem it solves is turning a user-supplied paper reference into readable text and rich, machine-usable metadata for summarizing, explaining, or analyzing the paper and its ecosystem.
The skill explains how to parse the arXiv paper ID from the various input forms, then fetch markdown via `curl` (either the `.md` endpoint or an `Accept: text/markdown` header, falling back to the HF page when no arXiv HTML exists). It documents the Hugging Face REST API endpoints: paper metadata as JSON, filters to find linked models/datasets/spaces by arXiv ID, an authenticated claim-authorship POST, the Daily Papers feed, paper listing with cursor pagination, hybrid semantic/full-text search, and paper indexing. Authenticated endpoints use a Bearer `HF_TOKEN`. It also notes error handling for papers not yet indexed (404) and the 14-day submission window for Daily Papers.
It targets researchers, ML engineers, and anyone who wants to summarize or explore AI papers along with their associated artifacts. Use cases include summarizing a shared paper, discovering the models and datasets a paper links to, browsing trending Daily Papers, and searching the corpus. Credentials are handled the standard way — an HF token passed as a bearer header — with no exfiltration.
A Hugging Face paper page URL, its .md variant, an arXiv abstract or PDF URL, or a bare arXiv ID (with or without a version suffix). It parses the paper ID and uses it against the API endpoints.
Fetch the markdown with curl from https://huggingface.co/papers/{PAPER_ID}.md, or request text/markdown via the Accept header on the normal paper URL. It relies on the arXiv HTML version, falling back to the HF page.
Endpoints like Daily Papers, listing, search, claiming authorship, and indexing use a Bearer token from the HF_TOKEN environment variable; fetching markdown and basic metadata generally do not.
The markdown endpoint returns a 404, meaning the paper is not yet indexed; the skill points to its error-handling and index-a-paper endpoint to add it.
Query the API with a filter of arxiv:{PAPER_ID} against the models, datasets, or spaces endpoints to list linked artifacts.
Quick Setup:
.claude/skills/Repository
huggingface/skills