arXiv Related Papers Downloader

What this skill does

• Accepts either:
  • topic + time range, or
  • seed paper (arXiv id or title)
• Finds related papers from arXiv API.
• Downloads PDF files into arxiv/.
• Uses filename format: versionDate-title.pdf, where versionDate is vN_YYYYMMDD.

When to use

Use this skill when the user asks to:

• crawl/search related papers by topic;
• find related papers from one article;
• download arXiv PDFs in batch;
• save with a deterministic naming rule.

Required user input

The user must provide one of these modes:

1. Topic mode

   • topic
   • start date (YYYY-MM-DD)
   • end date (YYYY-MM-DD)

2. Seed paper mode

   • seed arXiv id (preferred) or seed title
   • optional start date / end date

Optional:

• max results (default: 20)

Execution steps

1. Confirm missing parameters with the user.
2. Run the script from workspace root:

# Topic mode
python ./scripts/download_arxiv.py \
  --topic "graph neural network" \
  --start-date 2024-01-01 \
  --end-date 2024-12-31 \
  --max-results 20

# Seed mode by arXiv id
python ./scripts/download_arxiv.py \
  --seed-id "2401.12345v1" \
  --max-results 20

# Seed mode by title
python ./scripts/download_arxiv.py \
  --seed-title "Attention Is All You Need" \
  --start-date 2018-01-01 \
  --end-date 2024-12-31 \
  --max-results 20

3. Report back:
   • how many papers were found;
   • how many PDFs were downloaded;
   • the output directory path.

Output location and naming

• Output dir: ./arxiv/ (auto-created if missing)
• File naming rule:
  • v1_20240213-Your_Paper_Title.pdf
  • v3_20231105-Your_Paper_Title.pdf

Notes

• HTTPS uses normal TLS verification only (no insecure certificate bypass).
• The script only uses Python standard library.
• If a paper has no PDF link or download fails, it is skipped with a warning.
• Existing files are not downloaded again.