finding-open-access-papers_skill

Overview

This skill uses the Unpaywall API to locate legally available open access copies of papers that appear paywalled. It helps recover published versions, accepted manuscripts, or preprints so you can obtain full text without subscription access. It is designed to plug into a full-text fetching workflow and report source, version, and PDF URL.

How this skill works

Given a DOI, the skill queries the Unpaywall REST API (email parameter required) and inspects the JSON response fields like is_oa, best_oa_location, and oa_locations. If an open access location is found it prioritizes publishedVersion then acceptedVersion then submittedVersion, extracts a PDF or landing-page URL, and optionally downloads the file to a local folder. The skill includes simple rate-limiting, caching recommendations, and error handling for missing DOIs or network failures.

When to use it

• After DOI resolution triggers a paywall or subscription prompt
• When PubMed Central does not provide full text
• Before contacting authors or requesting interlibrary loan
• When a paper is highly relevant and you need the full text (e.g., relevance score ≥7)
• As part of an automated full-text fetching stage in a research pipeline

Best practices

• Ask the user for their real email at the start of the session and include it in all Unpaywall requests (do not use placeholders).
• Cache Unpaywall responses and add a small delay (e.g., 100 ms) between requests to be respectful of the service.
• Prefer best_oa_location.url_for_pdf when available; fall back to best_oa_location.url or entries in oa_locations.
• Treat publishedVersion as highest quality, then acceptedVersion, then submittedVersion; record version and host_type with downloads.
• Retry transient network errors with exponential backoff (max 3 attempts) and report failures clearly to the user.

Example use cases

• You attempted DOI resolution and hit a subscription wall; this skill finds a Europe PMC published PDF and downloads it.
• Automated pipeline checks for full text: PMC → DOI resolution → Unpaywall; if OA found, pipeline saves PDF and notes version.
• Manual research: user supplies a DOI and email; skill returns best OA URL, version, and host so researcher can access full text.
• Batch processing: script checks a shortlist of high-priority DOIs (score ≥7), caches results, and downloads available OA PDFs.