arXiv Research Skill

Overview

This skill enables systematic academic research through three core capabilities that form the minimal complete loop of knowledge building:

code

connect -> understand -> evidence
  Find  ->  Comprehend -> Cite

Setup

Install dependencies: uv sync

The Three Pillars

1. Connect (Knowledge Navigation)

Purpose: Find relevant existing knowledge

When to use:

•Starting research on a new topic
•Finding related work for a paper
•Discovering what exists in a field

Capabilities:

•Semantic search across arXiv
•Filter by category, author, date
•Rank by citation impact (via Semantic Scholar)
•Find similar papers to a known paper

Usage:

bash

uv run python scripts/connect.py search "transformer attention mechanism" --category cs.LG --limit 20
uv run python scripts/connect.py search "LLM agents" --since 2023-01 --until 2024-06  # Date filtering
uv run python scripts/connect.py similar "2301.00001" --limit 10
uv run python scripts/connect.py recent cs.AI --days 7
uv run python scripts/connect.py by-author "Yann LeCun"
uv run python scripts/connect.py cited-by "2301.00001" --limit 20  # Forward citations
uv run python scripts/connect.py coauthors "Yann LeCun" --limit 20  # Collaboration network

2. Understand (Meaning Extraction)

Purpose: Comprehend what the knowledge contains

When to use:

•Need to quickly grasp a paper's contribution
•Extracting methodology details
•Comparing multiple papers
•Writing literature review sections

Capabilities:

•Structured paper analysis (problem, method, contribution, limitations)
•Key findings extraction
•Methodology breakdown
•Multi-paper comparison

Usage:

bash

# Get paper content for analysis (single or batch)
uv run python scripts/connect.py content "2301.00001"
uv run python scripts/connect.py content "2301.00001,2302.00002,2303.00003"

# Pipe content into analysis prompts
uv run python scripts/connect.py content "2301.00001" | uv run python scripts/understand.py analyze quick

Available prompts: uv run python scripts/understand.py list

Prompts: quick, methodology, contribution, critical, compare, literature, implementation, evidence

3. Evidence (Source Attribution)

Purpose: Create verifiable links to sources

When to use:

•Writing academic papers
•Need proper citations
•Building bibliography
•Ensuring traceability of claims

Capabilities:

•BibTeX generation
•Multiple citation formats (APA, IEEE, ACM, Chicago, RIS)
•Batch citation export
•RIS export for Zotero/Mendeley/EndNote

Usage:

bash

uv run python scripts/evidence.py bibtex "2301.00001"
uv run python scripts/evidence.py apa "2301.00001"
uv run python scripts/evidence.py ris "2301.00001"  # For Zotero/Mendeley
uv run python scripts/evidence.py batch "2301.00001,2302.00002,2303.00003" --format bibtex
uv run python scripts/evidence.py batch "2301.00001,2302.00002" --format ris > refs.ris

4. TikZ (Figure Extraction)

Purpose: Extract TikZ source code from arXiv paper LaTeX sources

When to use:

•Reusing or adapting figures from papers
•Analyzing visualization techniques
•Understanding diagram construction

Capabilities:

•Extracts tikzpicture, tikzcd, circuitikz, pgfplots environments
•Captures captions, labels, and library dependencies
•Outputs as pure TikZ, compilable LaTeX, JSON, or brief summary

Usage:

bash

uv run python scripts/tikz.py extract "2301.00001"
uv run python scripts/tikz.py extract "2301.00001" --format latex > figures.tex
uv run python scripts/tikz.py extract "2301.00001,2302.00002" --format json
uv run python scripts/tikz.py list "2301.00001"
uv run python scripts/tikz.py extract "2301.00001" --format tikz | uv run python scripts/understand.py analyze quick

Workflow Examples

Literature Review Workflow

code

Progress:
- [ ] Step 1: Find seed papers
- [ ] Step 2: Expand with similar papers
- [ ] Step 3: Analyze each paper
- [ ] Step 4: Generate bibliography

bash

# Step 1: Find seed papers (ranked by citation impact)
uv run python scripts/connect.py search "your topic" --limit 50 --with-citations --sort citations

# Step 2: Expand with similar papers from top results
uv run python scripts/connect.py similar "top_paper_id"

# Step 3: Analyze each paper
uv run python scripts/connect.py content "paper_id" | uv run python scripts/understand.py analyze literature

# Step 4: Generate bibliography
uv run python scripts/evidence.py batch "id1,id2,id3" --format bibtex > refs.bib

Finding Evidence for a Claim

bash

# 1. Search for supporting research
uv run python scripts/connect.py search "your claim keywords" --with-citations

# 2. Verify the paper supports your claim
uv run python scripts/connect.py content "paper_id" | uv run python scripts/understand.py analyze evidence

# 3. Generate proper citation
uv run python scripts/evidence.py apa "paper_id"

API Dependencies

Service	Purpose	Rate Limit	API Key Required
arXiv	Paper search, content	1 req/3s	No
Semantic Scholar	Citations, similar papers	100 req/5min	No (optional for higher limits)
Jina Reader	Full text extraction	Generous	No

File Structure

code

arxiv-research-skill/
├── SKILL.md              # This file - usage guide
└── scripts/
    ├── connect.py        # Knowledge navigation
    ├── understand.py     # Analysis prompts
    ├── evidence.py       # Citation generation
    ├── tikz.py           # TikZ figure extraction
    ├── cache.py          # SQLite caching (~/.cache/arxiv-research/papers.db)
    └── utils.py          # Shared utilities (extractPaperId, cleanText)

Common Patterns

Finding Foundational Papers

bash

uv run python scripts/connect.py search "topic" --sort citations --limit 10

Tracking Recent Developments

bash

uv run python scripts/connect.py recent cs.AI --days 30

Building a Reading List

bash

uv run python scripts/connect.py search "topic" > papers.json
# Review and filter
uv run python scripts/evidence.py batch "selected_ids" --format bibtex

Error Handling

•Rate limited: Wait and retry, scripts have built-in backoff
•Paper not found: Verify arXiv ID format (YYMM.NNNNN)
•No citations: Paper may be too new for Semantic Scholar