Arxiv
This skill should be used when the user asks to "search arXiv", "find arXiv papers", "read arXiv paper", "get full text of arXiv", mentions an arXiv ID (YYMM.NNNNN format), references an arxiv.org URL, or asks about papers in CS, physics, math, or quantitative biology preprint categories (cs.CL, cs.AI, cs.LG, cs.CV, etc.).From its SKILL.md
npx -y skills add yuanhao96/pubmed-reader-cskill --skill arxivAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 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.
SKILL.md
5.5 KB, as published. Nobody here has run it
arXiv — Search, Read & Full Text
Use this skill for all arXiv operations: searching for preprints, reading paper abstracts by arXiv ID, and retrieving full HTML text.
When to Use This Skill
- Search: "Search arXiv for LLM memory", "Find arXiv papers on diffusion models"
- Read paper: "Read arXiv 1706.03762", "Get the paper at arxiv.org/abs/2301.12345"
- Full text: "Get full text of arXiv 2602.04557v1", "Read the full arXiv paper"
- Identifiers: Any query containing an arXiv ID (YYMM.NNNNN), arXiv URL, or "arxiv"
Keywords That Trigger Activation
arxiv, arXiv, search arxiv, read arxiv, arxiv paper, arxiv preprint, cs.CL, cs.AI, cs.LG, cs.CV, machine learning papers, AI papers, deep learning papers, NLP papers, computer science papers, physics papers, math papers, quantitative biology
Do NOT use for
- PubMed queries → use
pubmed-lookup - bioRxiv/medRxiv queries → use
biorxiv
Data Source: arXiv API
Search endpoint: https://export.arxiv.org/api/query
| Parameter | Description |
|---|---|
search_query | Terms with field prefixes: ti:, au:, abs:, cat:, all: |
id_list | Comma-separated arXiv IDs for direct lookup |
start | Pagination offset (0-based) |
max_results | Results per page (max 2000) |
sortBy | relevance, lastUpdatedDate, submittedDate |
HTML full text: https://arxiv.org/html/{id} (LaTeXML-rendered, not all papers)
Rate limits: 1 request per 3 seconds minimum. No authentication required.
Workflow 8: Search arXiv
User says: "Search arXiv for LLM memory augmentation"
Script: scripts/search_arxiv.py
Process:
- Parse query and extract terms
- Build arXiv query with field prefixes
- Call arXiv API
- Format results with arXiv citation format
Example:
from search_arxiv import search_arxiv, build_arxiv_query
# Basic search
results = search_arxiv("LLM memory", max_results=10)
# With category filter
results = search_arxiv("attention mechanism", category="cs.CL", max_results=10)
# By category
from search_arxiv import search_arxiv_by_category
results = search_arxiv_by_category("cs.AI", "reasoning", max_results=10)
Output:
## arXiv Search Results: "LLM memory augmentation"
Found 456 papers. Showing 10:
1. Smith A, Jones B, et al. Memory-Augmented Language Models. arXiv:2501.12345. 2025. [cs.CL].
Abstract: We propose a novel memory architecture...
2. Chen X, Wang Y, et al. Efficient Long-Term Memory in Transformers. arXiv:2412.67890. 2024. [cs.LG].
Abstract: This paper introduces...
Workflow 9: Read arXiv Paper
User says: "Read arXiv 1706.03762"
Script: scripts/fetch_arxiv.py
Process:
- Extract and validate arXiv ID (format: YYMM.NNNNN or YYMM.NNNNNvN)
- Call
fetch_arxiv_paper(arxiv_id) - Format as structured reference with abstract
Example:
from fetch_arxiv import fetch_arxiv_paper
result = fetch_arxiv_paper("1706.03762")
# Also accepts URLs: fetch_arxiv_paper("https://arxiv.org/abs/1706.03762")
Output:
## arXiv Paper: 1706.03762v7
**Reference**: Vaswani A, Shazeer N, Parmar N, et al. Attention Is All You Need. arXiv:1706.03762. 2017. [cs.CL].
**Abstract**:
The dominant sequence transduction models are based on complex recurrent or
convolutional neural networks...
**Categories**: cs.CL, cs.AI
**Comment**: 15 pages, 5 figures
**PDF**: https://arxiv.org/pdf/1706.03762
**HTML**: https://arxiv.org/html/1706.03762
Workflow 10: Get arXiv Full Text (HTML)
User says: "Get full text of arXiv 2602.04557v1"
Script: scripts/fetch_arxiv_fulltext.py
Process:
- Extract and validate arXiv ID
- Fetch HTML from
https://arxiv.org/html/{id} - Parse LaTeXML structure: sections, figures, references
- Return structured full text
Example:
from fetch_arxiv_fulltext import get_arxiv_fulltext, check_html_availability
avail = check_html_availability("2602.04557v1")
if avail:
result = get_arxiv_fulltext("2602.04557v1")
Output:
## Full Text: arXiv:2602.04557v1
**Title**: Paper Title Here
**Word Count**: 8,500 words
### Abstract
[...]
### Introduction
[...]
### Methods
[...]
### Results
[...]
### Conclusion
[...]
### Figures (5 total)
- fig1: Architecture overview...
- fig2: Performance comparison...
### References (42 total)
- Vaswani A, et al. Attention Is All You Need. NeurIPS 2017.
Note: Not all papers have HTML versions — only those processed by LaTeXML. If unavailable, direct user to the PDF.
Error Handling
| Error | Solution |
|---|---|
| arXiv ID not found | Verify format is YYMM.NNNNN |
| HTML not available | Direct user to PDF at arxiv.org/pdf/{id} |
| Rate limit | Wait 3 seconds between requests |
Available Scripts
scripts/search_arxiv.py—search_arxiv(query, max_results, category),build_arxiv_query(...)scripts/fetch_arxiv.py—fetch_arxiv_paper(arxiv_id),batch_fetch_arxiv(arxiv_ids)scripts/fetch_arxiv_fulltext.py—get_arxiv_fulltext(arxiv_id),check_html_availability(arxiv_id)scripts/utils/helpers.py—validate_arxiv_id(arxiv_id),format_arxiv_citation(article)