Arxiv Osiris

Search and download research papers from arXiv.org - Research version for OpenClaw agents

Security Scan

The skill's code and instructions are generally consistent with an arXiv search/download tool; issues are documentation mismatches and minor input-validation bugs rather than evidence of malicious behavior.

Purpose & Capability

Name/description claim: search and download from arXiv. The code uses the arxiv Python client to search and download PDFs and exposes the expected functionality. Required binary (python) and the SKILL.md's pip dependency (arxiv) align with the stated purpose.

Instruction Scope

SKILL.md describes searching, filtering, and downloading via a script and Python API only; it does not instruct reading unrelated files or exfiltrating data. However, documentation examples reference a PowerShell script named arxiv.ps1 and different CLI flags than the provided Python script (arxiv_search.py). That mismatch is a documentation/integration inconsistency and could confuse users or lead to running the wrong wrapper.

Install Mechanism

No install spec in registry (instruction-only); SKILL.md recommends pip install arxiv which is appropriate. There are no downloads from arbitrary URLs or extract steps. This is low-risk for installation.

Credentials

The skill requires no environment variables, no credentials, and no config paths. That is proportional to its function (public arXiv access). The code writes downloads into the user's Downloads/arxiv directory (uses HOME), which is expected for a download tool.

Persistence & Privilege

The skill is not forced always-on (always:false) and does not request elevated or persistent privileges. It does not modify other skills or global agent settings.

Assessment

This skill appears to do what it says: search arXiv and download PDFs using the Python arxiv package. Before installing or running it: 1) Be aware the SKILL.md examples reference a PowerShell script (arxiv.ps1) while the distributed file is a Python script (arxiv_search.py) — use the Python script or update wrappers accordingly. 2) The arXiv ID validation logic in the code is slightly odd and may reject some valid IDs; if you rely on automated downloads, review/patch is_valid_arxiv_id. 3) Run pip install arxiv from a controlled environment (virtualenv) and audit the pip package source if you have strict supply-chain requirements. 4) Because the skill will download PDFs into ~/Downloads/arxiv, consider running it in a sandbox or adjusting the download directory if you prefer a different location. If you need higher assurance, request provenance (homepage/source repo) from the publisher or inspect the code locally before use.

ArXiv Skill

Search and download scientific papers from arXiv.org - the largest free distribution of scientific preprints.

What it does

• Search papers by keywords, titles, abstracts
• Download PDFs directly
• Filter by category (physics, cs, math, etc.)
• Get metadata including authors, dates, categories

Installation

# Install Python dependency
pip install arxiv

Usage

Search for papers

# Basic search
.\arxiv.ps1 -Action search -Query "quantum computing"

# With max results
.\arxiv.ps1 -Action search -Query "machine learning" -MaxResults 10

# With category filter (physics, cs, math, q-bio, etc.)
.\arxiv.ps1 -Action search -Query "neural networks" -Categories "cs,stat"

Download a paper

# By arXiv ID
.\arxiv.ps1 -Action download -ArxivId "2310.12345"

Python API

from arxiv import search, download

# Search
results = search("simulation hypothesis", max_results=5)
for paper in results:
    print(f"{paper.title} - {paper.pdf_url}")

# Download
paper.download("/path/to/save")

Categories

Common arXiv categories:

• cs.* - Computer Science
• physics.* - Physics
• math.* - Mathematics
• q-bio.* - Quantitative Biology
• q-fin.* - Quantitative Finance
• stat.* - Statistics

Examples

• Search for consciousness papers: arxiv.ps1 -search "consciousness" -max 5
• Find physics papers: arxiv.ps1 -search "quantum" -cats "physics" -max 10
• Download paper: arxiv.ps1 -download "1706.03762" (Attention is All You Need)

Notes

• arXiv is free and open
• Papers are preprints - may not be peer-reviewed
• Great for staying current with research