Guide · Agent skill
How to use the ArXiv paper search agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The ArXiv skill is a packaged set of instructions and scripts that lets an AI agent search and read papers on arXiv, the open preprint server. It is for researchers, students and engineers who want to track new AI, machine learning, NLP or security papers, or get a quick read on one paper before deciding whether to read it in full. After this guide you will know what the skill does, how to install it, how to phrase requests so it triggers, and what to check before you cite anything it returns.
ArXiv Paper Searchdanielmiessler/LifeOS · MIT Download the skill Quick start
- 1Download the ZIP. Download the ArXiv skill ZIP from this page and unzip it on your computer.
- 2Place the folder. Move the unzipped folder into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the ZIP.
- 4Check requirements. Confirm the agent can reach export.arxiv.org and alphaxiv.org and that curl is available in its shell.
- 5Ask for latest papers. Request recent papers in a category, such as the latest cs.AI submissions on a topic.
- 6Run a topic search. Ask the agent to search arXiv for a keyword or phrase to get ranked recent matches.
- 7Summarize one paper. Give a paper ID or arXiv link and ask for a summary, then verify key claims against the PDF before citing.
What an agent skill is and what this one does
An agent skill is a folder with a SKILL.md file and supporting files that an AI agent loads when a request matches the skill's trigger words. The ArXiv skill covers three kinds of request, called intents: Latest, Search and Paper.
Latest lists recent submissions in one or more arXiv categories, sorted by submission date. Search runs a topic or keyword query. Paper explains one specific paper from its arXiv ID, a URL, or an AlphaXiv link.
- Latest: new submissions in categories such as cs.AI, cs.LG, cs.CL, cs.CR, cs.MA, cs.SE, cs.IR, cs.RO and cs.CV
- Search: topic, keyword or author queries against arXiv
- Paper: metadata from arXiv plus an AlphaXiv AI-generated overview when one exists
Install the ArXiv skill
Download the ArXiv skill as a ZIP from this page and unzip it. You can install it for yourself or for a single project.
For a personal install, unzip it into ~/.claude/skills/. For a project install, unzip it into the project's .claude/skills/ folder. Alternatively, if you use the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Requirements and what the skill needs
The skill needs network access to export.arxiv.org and alphaxiv.org, and curl available in the shell where the agent runs. It needs no API keys and no accounts.
The skill runs scripts and uses the network, so review the files before installing if that matters to you.
- Network access to export.arxiv.org and alphaxiv.org
- curl available in the shell
- No API keys or accounts
How to trigger the skill with natural requests
You do not need special syntax. Ask in plain language and include the topic, category or paper ID. The skill is triggered by words such as arxiv, papers, latest papers, research papers, recent ML papers, paper lookup, summarize paper, and cs.AI latest.
Name the intent when you can. Asking for the latest papers points to Latest, asking to search arXiv for a topic points to Search, and giving an ID or URL points to Paper.
- "what's new in AI safety papers this week" (Latest)
- "search arxiv for prompt injection defenses" (Search)
- "explain this paper: 2401.12345" (Paper)
Examples of useful requests
Each example below shows a request and what the agent returns. Results are ranked by relevance and each paper links back to arXiv.
For Latest, the agent lists new cs.AI submissions by their published date, with titles, authors, abstracts, links and a short pick of papers worth reading. For Search, it returns recent matches with abstracts and related searches. For Paper, it gives a summary from the AlphaXiv overview, or from the abstract if no overview exists, along with key contributions and the PDF link.
Safety notes and checking results
AlphaXiv overviews are AI-generated. The skill says they are useful for quick understanding but must be checked against the actual paper before you cite them. Use the arXiv link or PDF to confirm authors, dates and claims.
Check newness using the published date, not the last updated date, because updated dates change when old papers are edited. The arXiv API allows roughly one call every 3 seconds, so batch queries rather than repeating them.
- Verify any AlphaXiv summary against the paper before citing it
- Use the published date to judge whether a paper is new
- The skill's own scope excludes general web research, which belongs to a separate Research skill
Tips for better results
Name the categories or topic words you care about. The skill maps topic words such as LLMs, security or RAG to specific categories, and you can name a category directly.
If you have a preferences file for the skill at the customizations folder, it is used to rank results by your focus areas. Without one, relevance is judged from your request.
- Use the category name, such as cs.CR, when you want a narrow scan
- Give a paper ID without a version suffix; the skill strips it for the overview lookup
- Ask for a time window, such as this week, to keep results current
What people use it for
- Keeping up with new cs.AI or cs.LG submissions each week
- Searching arXiv for prompt injection or AI safety work before a literature review
- Getting a quick summary of a paper to decide whether to read it in full
- Building a regular scan of new cs.CL or cs.IR papers for a team
Frequently asked questions
How do I install the ArXiv agent skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ in a project. In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.
Does the ArXiv skill need an API key or account?
No. The skill uses the public arXiv API and AlphaXiv overview URLs, which need no authentication. It does need network access to export.arxiv.org and alphaxiv.org and curl in the shell.
Is the ArXiv skill free?
The source material for this page lists the skill under the MIT licence. Check the licence file in the repository for the terms that apply to your use.
Can I trust the AlphaXiv summaries?
Treat them as a starting point. The skill itself warns that AlphaXiv overviews are AI-generated and must be verified against the actual paper before you cite them.
Why does the skill say a paper is not new when it was updated?
The skill checks the published date, not the last updated date, because edits to older papers change the updated date. Only papers with a recent published date are treated as new submissions.
How is this different from the Research skill?
The ArXiv skill covers arXiv papers only. General web research is handled by a separate Research skill, and the ArXiv skill excludes URL parsing and annual reports.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use the ArXiv skill if you are a researcher, student or engineer who wants an agent to scan recent arXiv papers in AI, ML, NLP or security, or to summarise a paper before reading it. Pick something else if you need general web research, non-arXiv journals, or citation-ready summaries without checking the source paper yourself.
Download the skill