extract-paper-authors

Extract Paper Authors from a PDF

Given one paper PDF, return structured author data. You (Claude) read the PDF directly — no external service. Standalone use needs no plugin scripts (just curl + the Read tool), so the working directory does not matter here.

Steps

1. If given a URL (not a local path), download it first with Bash: curl -sL -o /tmp/paper.pdf "<url>" (or ask the user for the file). When the orchestrator calls this skill, the PDF is already in the run's cache/pdfs/.
2. Use the Read tool on the PDF (first 1–2 pages hold the author block).
3. Extract, in author order:
   • full_name (prefer the PDF's spelling; expand initials only if unambiguous)
   • affiliation(s) — the institution(s) tied to each author by footnote/superscript
   • country if printed
   • email if printed; else null
   • corresponding true/false if marked (✉, *, "corresponding author"); else null
4. Map affiliations correctly when authors share superscripts (1, 2, †). Do not guess — if a tie is ambiguous, attach all plausible institutions and note it.
5. Return JSON only:

{"authors": [
  {"position": 1, "full_name": "...", "affiliation": "...", "country": "US",
   "email": null, "corresponding": true}
]}

Rules

• Never invent an affiliation. If the PDF doesn't state one, use null.
• Scanned/image PDFs with no text layer: say so and return an empty authors list with "pdf_failed": true rather than hallucinating.