Skip to main content
skill CX Codex

arxiv-viewer — Claude Code skill: what it does, how to install

arxiv-viewer is a Claude Code skill for research workflows that searches, downloads, and reads academic papers from arXiv using its official API and ar5iv HTML pages.

Apache-2.0 133 installs ★ 1,632 updated 2026-10-19

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add actionbook/actionbook --skill arxiv-viewer -a claude-code
Open on GitHub

What arxiv-viewer does

This skill combines three separate endpoints for research workflows. Standard metadata lookups and field queries run through the arXiv API via WebFetch. Recent submissions and trending papers scrape arxiv.org through browser-fetcher, while full text inspection routes to ar5iv.org HTML pages via html-reader.

Instead of parsing raw PDF files for every query, the skill extracts sections, equations, tables, and references straight from HTML selectors. When you need the original document, it downloads the PDF file using curl.

Workflow rules require waiting for browser-fetcher or html-reader agents to finish completely before attempting any fallback requests.

What it helps with

  • Search arXiv by title, author, abstract, or category
  • Generate an AI overview report for any paper ID
  • Extract equations, figures, and citations from ar5iv HTML
  • Track recent submissions across computer science categories

Example prompts for the agent

  • /arxiv-viewer:search ti:transformer AND cat:cs.CL
  • /arxiv-viewer:report 2301.07041
  • /arxiv-viewer:figures 2301.07041

How to install arxiv-viewer

1Skills CLI (Vercel)

npx skills add actionbook/actionbook --skill arxiv-viewer -a claude-code

2Manually: put the files in place

~/.claude/skills/arxiv-viewer/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/arxiv-viewer/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about arxiv-viewer

How to install arxiv-viewer in Claude Code?

Run npx skills add actionbook/actionbook --skill arxiv-viewer -a claude-code in your terminal. The configuration is placed at ~/.claude/skills/arxiv-viewer/SKILL.md.

What does arxiv-viewer do?

It lets your coding agent query arXiv metadata, fetch abstracts, read specific paper sections from ar5iv HTML, and download source PDFs.

Why does the skill use ar5iv.org?

ar5iv converts arXiv papers into responsive HTML. The agent parses concrete CSS selectors for sections, math equations, and bibliography without downloading the whole PDF.

Source text

SKILL.md · Apache-2.0
---
name: arxiv-viewer
description: View, search, and download academic papers from arXiv. Supports API queries, web scraping via Actionbook, and HTML paper reading via ar5iv. Use when user asks about arxiv papers, academic papers, research papers, paper summaries, latest papers, or wants to search/download/read papers.
---

# arXiv Viewer

Access, search, download, and read academic papers from arXiv using a hybrid API + Actionbook approach.

## Architecture Overview

```
┌─────────────────────────────────────────────────────────────┐
│                     arxiv-viewer                            │
├─────────────────┬─────────────────┬─────────────────────────┤
│   arXiv API     │  arxiv.org Web  │      ar5iv.org          │
│   (WebFetch)    │  (Actionbook)   │    (Actionbook)         │
├─────────────────┼─────────────────┼─────────────────────────┤
│ • Metadata      │ • Latest list   │ • Read sections         │
│ • Search        │ • Trending      │ • Extract figures       │
│ • By ID lookup  │ • Advanced      │ • Extract citations     │
│                 │   search form   │ • Get outline           │
└─────────────────┴─────────────────┴─────────────────────────┘
```

## Tool Priority

**By Feature:**

| Feature | Primary Tool | Fallback |
|---------|--------------|----------|
| Paper metadata | WebFetch (API) | browser-fetcher |
| Search | WebFetch (API) | browser-fetcher |
| Latest papers | browser-fetcher (Actionbook) | WebFetch (API) |
| Trending | browser-fetcher (Actionbook) | - |
| Advanced search | browser-fetcher (Actionbook) | WebFetch (API) |
| Read HTML section | html-reader (Actionbook) | Read (PDF) |
| Download PDF | Bash (curl) | - |

## Workflow Rules

### ⚠️ Agent Waiting Rule

**After launching browser-fetcher or html-reader agents:**
1. ✅ **MUST wait for ALL agents to complete**
2. ⛔ **DO NOT** use WebFetch/WebSearch while waiting
3. ✅ Only use fallback tools after agents have failed

---

## Data Sources

### 1. arXiv API (WebFetch)

**Best for:** Quick metadata lookup, simple search

```
Base URL: http://export.arxiv.org/api/query
```

| Parameter | Description |
|-----------|-------------|
| search_query | Search with field prefixes (ti:, au:, abs:, cat:) |
| id_list | Comma-separated arXiv IDs |
| max_results | 1-2000 (default: 10) |
| sortBy | relevance / submittedDate / lastUpdatedDate |

**Field Prefixes:** `ti:` (title), `au:` (author), `abs:` (abstract), `cat:` (category), `all:` (all)

**Boolean Operators:** `AND`, `OR`, `ANDNOT` (UPPERCASE)

### 2. arxiv.org Web (Actionbook + agent-browser)

**Best for:** Latest papers, trending, advanced search UI

| Page | Action ID | Use Case |
|------|-----------|----------|
| Latest list | `arxiv.org/list/{category}/recent` | Recent submissions |
| Advanced search | `arxiv.org/search/advanced` | Complex filters |
| Homepage | `arxiv.org/` | Trending/announcements |

**Workflow:**
```
1. search_actions("arxiv list recent")
2. get_action_by_id(action_id) → selectors
3. agent-browser open URL
4. agent-browser get text <selector>
5. Return results
```

### 3. ar5iv.org HTML Papers (Actionbook + agent-browser)

**Best for:** Reading specific sections, extracting figures/citations

```
HTML Paper URL: https://ar5iv.org/html/{arxiv_id}
```

| Element | Selector | Description |
|---------|----------|-------------|
| Title | `.ltx_title` | Paper title |
| Authors | `.ltx_authors` | Author list |
| Abstract | `.ltx_abstract` | Abstract text |
| Sections | `section` | All sections |
| Section title | `h2.ltx_title`, `h3.ltx_title` | Section headings |
| Paragraphs | `.ltx_para` | Paragraph content |
| Figures | `figure.ltx_figure` | Figures with captions |
| Tables | `table.ltx_tabular` | Data tables |
| Equations | `.ltx_equation` | Math equations |
| Bibliography | `.ltx_bibliography` | Reference list |
| Single citation | `.ltx_bibitem` | Individual reference |

**Workflow:**
```
1. search_actions("ar5iv section")
2. get_action_by_id(action_id) → selectors
3. agent-browser open ar5iv.org/html/{id}
4. agent-browser get text <section_selector>
5. Return section content
```

---

## URL Patterns

| Purpose | URL |
|---------|-----|
| arXiv Abstract | `https://arxiv.org/abs/{id}` |
| arXiv PDF | `https://arxiv.org/pdf/{id}.pdf` |
| arXiv API | `http://export.arxiv.org/api/query?id_list={id}` |
| ar5iv HTML | `https://ar5iv.org/html/{id}` |
| ar5iv Abstract | `https://ar5iv.org/abs/{id}` |

## arXiv ID Formats

| Format | Example |
|--------|---------|
| New (2007+) | `2301.07041` |
| With version | `2301.07041v2` |
| Old | `cs.AI/0612345` |

## Common Categories

| Code | Field |
|------|-------|
| `cs.AI` | Artificial Intelligence |
| `cs.CL` | Computation and Language (NLP) |
| `cs.CV` | Computer Vision |
| `cs.LG` | Machine Learning |
| `cs.SE` | Software Engineering |
| `stat.ML` | Statistical ML |

---

## Feature Matrix

| Command | Data Source | Agent |
|---------|-------------|-------|
| `/arxiv-viewer:paper` | API | paper-fetcher |
| `/arxiv-viewer:search` | API | search-executor |
| `/arxiv-viewer:download` | Direct URL | - |
| `/arxiv-viewer:latest` | arxiv.org | browser-fetcher |
| `/arxiv-viewer:trending` | arxiv.org | browser-fetcher |
| `/arxiv-viewer:read` | ar5iv.org | html-reader |
| `/arxiv-viewer:outline` | ar5iv.org | html-reader |
| `/arxiv-viewer:figures` | ar5iv.org | html-reader |
| `/arxiv-viewer:citations` | ar5iv.org | html-reader |
| `/arxiv-viewer:report` | API + ar5iv | paper-summarizer |

---

## Output Formats

### Paper Info
```
## {Title}

**arXiv:** {id}
**Authors:** {author1}, {author2}, ...
**Categories:** {cat1}, {cat2}
**Published:** {date}

### Abstract
{abstract}

**Links:** [Abstract]({abs_url}) | [PDF]({pdf_url}) | [HTML]({ar5iv_url})
```

### Section Content
```
## {Section Title}

{section content}

---
*Source: ar5iv.org/html/{id}*
```

### Paper Report (AI Generated)

**Command:** `/arxiv-viewer:report {arxiv_id}` or `/arxiv-viewer:report {paper_title}`

**Purpose:** Generate a comprehensive, well-formatted paper report with AI-generated analysis.

**Output Format:**
```markdown
---
> **🤖 AI Generated Content**
> Author: Powered by ActionBook
---

# {Paper Title}

**Paper Information**
| Field | Content |
|-------|---------|
| arXiv ID | {id} |
| Authors | {authors} |
| Affiliations | {affiliations} |
| Published | {date} |
| Categories | {categories} |

---

## 📋 Abstract

{abstract_summary}

---

## 🎯 Problem Statement

{problem_statement}

---

## 💡 Key Contributions

1. {contribution_1}
2. {contribution_2}
3. {contribution_3}

---

## 🔬 Method Overview

{method_summary}

---

## 📊 Experimental Results

{experimental_results}

---

## 🌟 Why It Matters

{significance}

---

## 🔗 Links

- [arXiv Abstract](https://arxiv.org/abs/{id})
- [PDF Download](https://arxiv.org/pdf/{id}.pdf)
- [HTML Version](https://ar5iv.org/html/{id})

---
> This report was automatically generated by AI based on the original paper content.
> **Powered by ActionBook** | Generated: {timestamp}
```

**Workflow:**
```
1. Fetch paper metadata via arXiv API
2. Read full paper content from ar5iv.org HTML
3. Extract key sections (abstract, intro, method, results)
4. Generate structured report with AI analysis
5. Add ActionBook branding header and footer
```

**Report Sections:**
| Section | Source | Description |
|---------|--------|-------------|
| Paper Information | API metadata | Basic paper information |
| Abstract | Abstract | Summary of abstract |
| Problem Statement | Introduction | Problem statement extracted |
| Key Contributions | Introduction | Key contributions listed |
| Method Overview | Method section | Technical approach summary |
| Experimental Results | Experiments | Key results and metrics |
| Why It Matters | Analysis | AI-generated significance |

## Rate Limiting

- arXiv API: 3 second delay between requests
- ar5iv.org: Respect server load
- agent-browser: One page at a time

© actionbook · Apache-2.0 · actionbook/actionbook

Similar skills

All skills for Science and research →

From the same set9

Project sponsors

Companies that keep this analytics open