---

name: arxiv description: Use when the user mentions 'arxiv' or asks to find, search for, or download arxiv papers (by arxiv ID, author name, or topic), get paper source code, or retrieve LaTeX files from arxiv. Can search for papers to find arxiv IDs when only author/topic is provided, then download them. Prefer source format over PDF.

ArXiv Search and Download Skill

Download and view arXiv paper sources using the arxiv command-line tool.

Requires: requests (pip install requests)

ArXiv ID Format

• Modern: YYMM.NNNNN (e.g., 2301.07041)
• Old format: category/YYMMNNN (e.g., astro-ph/0701123)

Downloading Papers

Use the arxiv CLI tool bundled with this skill:

Copy# View all LaTeX source files (.tex, .bib, .bbl) with syntax highlighting
python scripts/arxiv.py 2301.07041

# Save entire source to a directory
python scripts/arxiv.py 2301.07041 --save

# List files in the archive
python scripts/arxiv.py 2301.07041 --list

# View only specific file types
python scripts/arxiv.py 2301.07041 --tex   # Only .tex files
python scripts/arxiv.py 2301.07041 --bib   # Only .bib files
python scripts/arxiv.py 2301.07041 --bbl   # Only .bbl files

The default output (no flags) wraps each file in markdown code blocks with appropriate language tags (latex/bibtex).

Searching ArXiv

Copy# Search by author
python scripts/arxiv.py --search "au:Handley"

# Search by title
python scripts/arxiv.py --search "ti:cosmology"

# Search by category
python scripts/arxiv.py --search "cat:astro-ph.CO"

# Combined search
python scripts/arxiv.py --search "ti:cosmology+AND+au:Planck"

# Limit results (default 10)
python scripts/arxiv.py --search "au:Hinton" -n 5

Search Query Prefixes

Typical Workflow

1. Search for papers: python scripts/arxiv.py --search "au:Author"
2. Pick an arxiv ID from results
3. View the source: python scripts/arxiv.py <id>
4. Use --save to extract files locally if needed

Integration with LLM Analysis

After saving source with --save:

Copypython scripts/arxiv.py 2301.07041 --save
code2prompt 2301.07041 --include "*.tex" --output-file /tmp/paper.md

Then pass to external LLMs for analysis.