Paywall Strategist

Paywall Strategist — The Access Negotiator

Half the high-quality biotech literature is paywalled. The same is true for legal databases, financial research, and many institutional reports. Tagging "Unverifiable — paywalled" without first checking for legitimate open-access mirrors is lazy research and violates Guiding Principle #4 (trace to primary sources).

This skill exists so Spelunker can honestly say "I tried these specific avenues" before giving up.

Guiding Principles

1. Open-access first. Most published research has a legitimate open mirror. Find it before declaring defeat.
2. Use legal channels only. No Sci-Hub, no Library Genesis, no shadow libraries. The skill helps users find papers they can legitimately access; it does not bypass copyright.
3. An abstract is a partial source, not a non-source. A well-extracted abstract often answers ~70% of the verification question.
4. Ask the author when stuck. Most academic authors will share their own paper on request. The cost is one email.

How to Run

Input

From the Spelunker orchestrator (typically via source-triangulator hitting a paywall):

• The paywalled URL or DOI
• The atomic claim being verified
• Optional: the user's institutional affiliation (some users have library access we should suggest)

Steps

Step 1 — Identify the Source

Extract the canonical identifier:

• DOI → use as primary key (DOIs resolve to authoritative metadata)
• arXiv ID, SSRN ID, NBER number → use as primary key
• URL only → fetch the page, extract title/authors/year, then search for canonical IDs

Record: title, authors, year, journal/venue, DOI (if any).

Step 2 — Search Open-Access Mirrors

Run these searches in order. Stop at the first hit, but record what was tried.

For biomedical/life-sciences claims:

1. unpaywall.org/<DOI> — checks ~30M open-access copies
2. europepmc.org/article/MED/<PMID> — Europe PMC mirrors many NIH-funded papers
3. biorxiv.org / medrxiv.org search by title — preprints often free
4. pubmedcentral.gov (PMC) — NIH open-access archive
5. Author's institutional repository (search " repository")

For finance/economics claims:

1. ssrn.com search by title — most working papers are free
2. nber.org/papers — NBER working papers
3. repec.org (RePEc) — economics archive
4. Federal Reserve / IMF / World Bank publication search if the paper is policy-adjacent

For tech/CS/AI claims:

1. arxiv.org search by title or author
2. semanticscholar.org — often surfaces the open PDF in the right sidebar
3. paperswithcode.com — for ML papers, may have the PDF link
4. Author's personal page or GitHub Pages site

For all domains:

1. Google Scholar for the title — the right-side "[PDF]" link is the legal open copy when present
2. OpenAlex (api.openalex.org/works/doi:) — returns open_access.is_oa and oa_url when available

Step 3 — Extract What's Available

Even when the full text remains paywalled:

1. Fetch and extract the abstract — almost always free.
2. Extract the figures and tables if available via the publisher's "free preview" mode.
3. Extract the methods section if the journal uses methods-open-access (some Nature/Science articles do).
4. Note the citations to/from this work via OpenCitations or Semantic Scholar — adjacent open-access papers may discuss the same claim.

Step 4 — Find Adjacent Open-Access Work

Search for:

• Preprints by the same authors on the same topic (often a non-paywalled draft of the same finding)
• Systematic reviews or meta-analyses that include this paper (the review's discussion summarizes the finding)
• Conference papers vs. journal papers — if the journal is paywalled, the conference version may be free
• Replication studies (e.g., on osf.io)

These are not substitutes for the original, but they may corroborate or contextualize.

Step 5 — Generate Ask-the-Author Template

If the paper truly cannot be accessed and the claim is critical, generate a polite request the user can send to the corresponding author. Template:

Subject: Request for PDF — "<paper title>"

Dear Dr. <last name>,

I'm researching <one-sentence claim/topic> and your paper "<title>" (<journal>, <year>) appears to be the most directly relevant primary source. The published version is behind a paywall I don't currently have access to.

Would you be willing to share a copy of the manuscript? I'm happy to share what I'm working on if helpful.

Thank you for considering,
<user name>

Surface the corresponding author's email (usually in the abstract or first page of the paper, or on their institutional page).

Output

PAYWALL STRATEGIST RESULT
─────────────────────────
Original source: <title>, <authors>, <year>, DOI: <doi>

Open-access mirror found: yes | no
  If yes: <URL>, mirror type (preprint / institutional / OA journal / unpaywall)
  If no: searched [list of mirrors], none available

Available extractions:
  - Abstract: [text or "not extracted"]
  - Methods: [text or "not extracted"]
  - Figures/tables: [N extracted, or "not available"]

Adjacent open-access work:
  - [Title] (citation): [why relevant — same authors / cites this / replication / etc.]
  - ...

Author contact (if needed):
  - Corresponding: <name>, <email>
  - Email template: see above

RECOMMENDATION:
  [verifiable_from_extractions | partially_verifiable_from_adjacent | author_outreach_recommended | truly_unverifiable]

If the claim is verifiable from the extractions or adjacent work, hand back to source-triangulator with the new evidence so the bundle can be updated. If author outreach is the only path, the orchestrator should escalate to Reentry Protocol Level 4 (user assist).

Error Handling

DOI doesn't resolve: The paper may have been retracted or merged. Check retractionwatch.com and surface this — a retracted source is itself important evidence.

Unpaywall / OpenAlex API down: Fall back to manual Google Scholar search.

No open mirror exists, no adjacent work, no author contact: Hand back to Spelunker with truly_unverifiable recommendation. Document everything tried so the Gaps section is honest.

Scope Boundaries

Does NOT: Bypass paywalls illegally. No Sci-Hub, no shared institutional credentials, no "12ft.io"-style scrapers. Does NOT: Replace primary access. If a claim is critical, the user should still try to access the paper directly through their institution. Does NOT: Vouch for preprints' rigor. Preprints lack peer review and should be tagged accordingly in the evidence bundle.