Paper Download Skill

Overview

• •This Skill helps you quickly download a paper PDF by a single DOI, or batch download PDFs for multiple DOIs.
• •It leverages the auto_paper_download package and provides two ready-to-run scripts.
• •PDFs are saved under downloads/pdfs/<doi-slug>/ with supplementary PDFs (if found) saved next to the main PDF.

Prerequisites

• •Python environment set up for this project (e.g., uv sync).
• •A .env file in the project root (copy from .env.example) with any credentials you have:
  • •WILEY_TDM_TOKEN (Wiley TDM API)
  • •ELSEVIER_API_KEY (Elsevier TDM API)
  • •SPRINGER_API_KEY (optional, open-access only)
  • •CROSSREF_MAILTO, OPENALEX_MAILTO (contact email for polite API usage)
  • •UNPAYWALL_EMAIL (optional, enables OA fallback)
  • •CROSSREF_REQUEST_DELAY, WILEY_REQUEST_DELAY (optional throttling)
• •Missing credentials simply disable that provider. Provide at least one CROSSREF_MAILTO or OPENALEX_MAILTO.

Scripts

• •scripts/download_by_doi.py: download a single DOI.
• •scripts/download_multiple_dois.py: download multiple DOIs (via repeated flags or a file).

DOI Examples and Templates

• •example_dois.txt: Ready-to-use example DOI file for testing.

See DOI_EXAMPLES.md for:

• •Valid DOI formats (standard and URL forms)
• •Publisher-specific DOI examples
• •File naming conventions for batch downloads
• •Complete usage examples and best practices

Single DOI Usage

Run from the project root:

python .claude/skills/paper-download/scripts/download_by_doi.py --doi 10.1038/s41586-020-2649-2 --verbose

Options:

• •--output-dir destination root, defaults to downloads/pdfs
• •--delay throttle seconds, default 1.5 (minimum 1.0)
• •--overwrite re-download even if exists
• •--dry-run inspect routing without downloading
• •--verbose debug logs

Multiple DOIs Usage

Provide DOIs directly or via a text file (one per line):

# Multiple DOIs via repeated flags
python .claude/skills/paper-download/scripts/download_multiple_dois.py \
  --doi 10.1038/s41586-020-2649-2 \
  --doi 10.1002/anie.202100001 \
  --verbose

# From a file of DOIs
python .claude/skills/paper-download/scripts/download_multiple_dois.py --doi-file ./dois.txt --delay 1.5

Options:

• •--doi repeatable flag to add DOIs
• •--doi-file path to a file with one DOI per line
• •--output-dir, --delay, --max-per-publisher, --overwrite, --dry-run, --verbose

Resume and Batching

For large runs, you can resume from a checkpoint and/or run in batches:

# Resume from the last checkpoint (derived from --doi-file name)
python .claude/skills/paper-download/scripts/download_multiple_dois.py \
  --doi-file ./dois.txt \
  --resume \
  --delay 1.5 --verbose

# Resume with a custom checkpoint file
python .claude/skills/paper-download/scripts/download_multiple_dois.py \
  --doi-file ./dois.txt \
  --resume --checkpoint-file downloads/state/dois.checkpoint.json \
  --delay 1.5

# Batch execution: process 500 DOIs per run
# Run batch index 0, then 1, etc.
python .claude/skills/paper-download/scripts/download_multiple_dois.py \
  --doi-file ./dois.txt --batch-size 500 --batch-index 0 --delay 1.5
python .claude/skills/paper-download/scripts/download_multiple_dois.py \
  --doi-file ./dois.txt --batch-size 500 --batch-index 1 --delay 1.5

Reports and checkpoints:

• •Checkpoints are stored under downloads/state/ by default (derived from --doi-file name).
• •Successes report: downloads/state/<name>_successes.txt (tab-separated DOI and saved path).
• •Failures report: downloads/state/<name>_failures.txt (tab-separated DOI and error or NO_OUTPUT).
• •Dry-run does not write checkpoints or reports.

Behavior Notes

• •The scripts automatically read .env. Missing providers are skipped gracefully.
• •When publisher/Crossref/OpenAlex cannot serve a PDF, Unpaywall OA fallback is attempted if UNPAYWALL_EMAIL is set.
• •Springer only returns open-access items; paywalled content still requires manual access.
• •After downloading a PDF, a DOI landing page scan looks for supplementary links and saves PDF-only assets.
• •Throttling ensures compliance with typical TDM limits (min 1.0s/file).

Troubleshooting

• •403/429 responses usually indicate rate limits or missing safelisting; use request delays and ensure credentials.
• •Check logs for the exact URL that failed when extending to new publishers.