Skip to content
Get packs

Guide · Agent skill

How to use the DeepXiv paper search and reading agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The deepxiv skill lets an AI agent search open-access academic papers through the DeepXiv service and read them in stages, starting with a short brief and only moving to the full text when you need it. It is aimed at researchers doing a literature review and students who want a quick summary followed by a specific section. By the end of this guide you will know what the skill does, how to install it, how to trigger it with plain requests, and what it needs to run.

DeepXiv Paper Search & Readingwanshuiyin/Auto-claude-code-research-in-sleep · MIT

Quick start

  1. 1Download and unzip the skill. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings, then Capabilities, then Skills.
  2. 2Install the DeepXiv package. Run pip install deepxiv-sdk with Python 3 installed.
  3. 3Make a search request. Ask your agent for papers on a topic, for example agent memory, and say how many results you want.
  4. 4Get a brief. Give the agent an arXiv ID. With no mode specified it returns the brief view.
  5. 5Read one section. Ask for a named section such as Introduction to read only that part of the paper.
  6. 6Check trending papers. Ask what is trending and give a window of 7, 14 or 30 days.

What the deepxiv skill makes the agent do

An agent skill is a folder with a SKILL.md file that tells an agent how to handle a certain kind of task. The deepxiv skill covers literature retrieval. Once it is installed, the agent can search open-access papers through DeepXiv and read them in layers instead of loading whole papers into the conversation at once.

The layers run in this order: search, then paper-brief, then paper-head, then paper-section, and the full paper only when you explicitly need it. The skill can also list trending papers and look up Semantic Scholar IDs for published venue metadata and citation counts.

  • search: find papers on a topic
  • paper-brief: a short summary of one paper
  • paper-head: the header or TLDR of one paper
  • paper-section: one named section, such as the Introduction
  • trending: recently trending papers, over 7, 14 or 30 days
  • sc: Semantic Scholar lookup for venue metadata and citation counts

Install the deepxiv skill

Download the skill ZIP from this page and unzip it. For Claude Code, place the unzipped skill folder in ~/.claude/skills/ for all your projects, or in a project's .claude/skills/ folder for one project only.

In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.

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

Install the DeepXiv Python package

DeepXiv is optional, but the skill needs a working way to reach the service. The skill's setup uses the deepxiv-sdk package, which requires Python 3. Install it with pip.

The skill's adapter script, deepxiv_fetch.py, comes from the ARIS repository. If you do not have that adapter, the skill falls back to the raw deepxiv command-line tool.

pip install deepxiv-sdk

How authentication works

Authentication is handled by the deepxiv-sdk command-line tool using a DEEPXIV_TOKEN value. If no token is configured, a command that needs one registers a token automatically and saves it to ~/.env. You do not need to create or paste a token yourself for this to work.

If python-dotenv is installed, the CLI also loads ~/.env and the current project's .env file on startup. Setting DEEPXIV_TOKEN in the CLI's environment skips automatic registration and token saving, but it does not stop the file loading. The skill itself does not read these files or display tokens.

Trigger the skill with natural requests

You do not have to memorise command syntax. Describe what you want and name the paper topic or arXiv ID. The skill recognises requests for layered paper access, section-level reading, trending papers, or DeepXiv-backed literature retrieval.

You can also add options directly to your request or to the slash command. The documented options are a maximum result count, a brief, head or section mode, trending with a day window, a web search, and a Semantic Scholar lookup.

  • "Search DeepXiv for recent papers on agent memory, give me five."
  • "Give me a brief of arXiv 2409.05591."
  • "Show the Introduction of arXiv 2409.05591."
  • "What papers are trending in the last 14 days?"
  • "Look up Semantic Scholar ID 258001 for venue and citation count."
/deepxiv "agent memory" - max: 5

Read papers in layers

Start broad and go deeper only when a paper looks useful. The skill is designed to avoid pulling a full paper into context. If you give it an arXiv ID with no mode, it defaults to the brief view.

When results come back, the agent presents a compact table for searches, and for paper reads it summarises the title, authors, date, TLDR and the suggested next depth step.

  • Search to shortlist papers
  • Brief to decide whether a paper is relevant
  • Head or section to read only the part you need
  • Full text only when you explicitly ask for it
/deepxiv "2409.05591" - section: Introduction

Safety notes: scripts and network use

This skill runs scripts and uses the network. Queries and paper IDs are sent to DeepXiv, so do not use it for work that must avoid third-party services. If DeepXiv is not installed, the skill tells you to install it and suggests the /arxiv skill or a web-based research search instead.

The skill is an additive source. It does not replace existing ARIS literature tooling. When a result overlaps with a published venue paper from Semantic Scholar, the richer venue metadata is kept in the final summary.

  • Runs local scripts (deepxiv_fetch.py or the deepxiv CLI)
  • Sends search terms and paper IDs to the DeepXiv service
  • Authentication is left to the CLI; the skill does not handle tokens

What people use it for

  • Building a shortlist of papers for a literature review by searching and reading briefs before reading any in depth.
  • Getting the Introduction or another named section of a paper without reading the full text.
  • Finding recently trending papers in a field over a chosen day window.
  • Looking up Semantic Scholar venue metadata and citation counts for a paper found through DeepXiv.

Frequently asked questions

Is the deepxiv skill free?

The source does not state pricing or licence terms for the DeepXiv service. The skill itself is listed under the MIT licence on this page, and the deepxiv-sdk package is a third-party tool that the skill uses.

What do I need to install before using deepxiv?

You need Python 3 and the deepxiv-sdk package, installed with pip install deepxiv-sdk. You also need either the ARIS repository's deepxiv_fetch.py adapter or the raw deepxiv CLI, which the skill uses as a fallback.

Do I need to create a DeepXiv token myself?

Not necessarily. The SDK can auto-register a DEEPXIV_TOKEN when a command needs one and save it to ~/.env. You can also set DEEPXIV_TOKEN in the environment yourself, which skips automatic registration.

Where does the deepxiv skill get installed?

For Claude Code, unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, upload the ZIP under Settings, then Capabilities, then Skills.

When should I use arXiv instead of deepxiv?

Use the /arxiv skill for simple arXiv listings, batch search, or PDF download. Use deepxiv when you want to inspect papers step by step, from brief to a named section.

Does deepxiv send my data anywhere?

Yes. The skill uses the network, and search queries and paper IDs go to DeepXiv. Avoid it for work that must not involve third-party services.

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 deepxiv skill if you review literature and want to skim many papers, reading only the briefs, headers or sections that matter. Skip it if you only need a plain arXiv listing or PDF download, where /arxiv is more direct, or if your work cannot send queries to a third-party service.

Download the skill