Arxiv search

Skill augchan42/inkstone/skills/arxiv-search

硯台 — Claude Code skills for creative and design workflows. Ink painting prompts from classical Chinese verse, bilingual verse explanations, and UI/UX auditing.

Install

npx -y skills add augchan42/inkstone --skill arxiv-search

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Search arXiv for papers by topic, author, or ID. Use when finding research, surveying recent publications, or looking up a specific paper.

SKILL.md

4.0 KB, as published. Nobody here has run it

arXiv Paper Search

Search the arXiv API for academic papers. No authentication required.

When to Use

User wants to find papers on a topic
User asks about recent research in a field
User provides an arXiv ID and wants details
User wants to compare papers or find related work

API Details

Base URL: https://export.arxiv.org/api/query
Auth: None required
Rate limit: 1 request per 3 seconds (be polite)
Response format: Atom XML

How to Search

By topic/keywords

curl -sL "https://export.arxiv.org/api/query?search_query=all:QUERY&max_results=10&sortBy=submittedDate&sortOrder=descending"

Replace spaces in QUERY with +. For multi-word phrases, use %22phrase%22 for exact match.

By specific field

ti: — title
au: — author
abs: — abstract
cat: — category (e.g. cs.AI, cs.CL, stat.ML)
all: — all fields

Combine with AND, OR, ANDNOT:

# Papers about transformers in NLP
curl -sL "https://export.arxiv.org/api/query?search_query=ti:transformer+AND+cat:cs.CL&max_results=10&sortBy=submittedDate&sortOrder=descending"

# Papers by a specific author
curl -sL "https://export.arxiv.org/api/query?search_query=au:hinton&max_results=10&sortBy=submittedDate&sortOrder=descending"

By arXiv ID

curl -sL "https://export.arxiv.org/api/query?id_list=2301.07041"

Multiple IDs: id_list=2301.07041,2305.18290

Pagination

Use start parameter:

# Results 11-20
curl -sL "https://export.arxiv.org/api/query?search_query=all:QUERY&start=10&max_results=10&sortBy=submittedDate&sortOrder=descending"

Parsing the Response

The response is Atom XML. Extract these fields from each <entry>:

Field	XML Path	Notes
Title	`<title>`	May contain line breaks — normalize
Authors	`<author><name>`	Multiple entries
Abstract	`<summary>`	May contain LaTeX
Published	`<published>`	ISO 8601 date
Updated	`<updated>`	Latest version date
arXiv ID	`<id>`	URL form, extract ID from path
PDF link	`<link rel="related" title="pdf">`	Direct PDF URL
Categories	`<category term="...">`	Primary + secondary
Comment	`<arxiv:comment>`	Page count, conference info

Output Format

Present results as a numbered list:

1. **Title of Paper** (arXiv:2301.07041)
   Authors: First Author, Second Author, et al.
   Published: 2023-01-17 | Categories: cs.CL, cs.AI
   Abstract: First 2-3 sentences of the abstract...
   PDF: https://arxiv.org/pdf/2301.07041

For single paper lookups, show the full abstract.

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
cs.CR	Cryptography and Security
stat.ML	Machine Learning (Stats)
math.OC	Optimization and Control
eess.SP	Signal Processing
quant-ph	Quantum Physics

Tips

Default to sortBy=submittedDate&sortOrder=descending for recent papers
Use max_results=10 as default, increase if user wants more
For broad topics, search by category + keywords to reduce noise
arXiv IDs after 2007 use the format YYMM.NNNNN (e.g. 2301.07041)
Older IDs use subject/YYMMNNN format (e.g. hep-th/9901001)
The abstract page is at https://arxiv.org/abs/ID
The PDF is at https://arxiv.org/pdf/ID

Reading Full Papers

If the user wants to read a paper's full content, fetch the PDF URL and use the Read tool on it, or use WebFetch on the abstract page for the HTML version.

For the abstract page (lighter weight):

curl -sL "https://arxiv.org/abs/2301.07041"