SuperAPI AI Directory

academic-reference-finder

Coding & development

View repository

Updated 10-15 23:58

About this skill

Academic Reference Finder

Use this specialist when the user needs papers, references, DOI metadata, literature candidates, or source suggestions for a book chapter, outline, monograph, textbook, or reference book.

Default policy: use no-key sources first. Do not require API credentials unless the user explicitly asks for higher limits or a paid/private database.

No-Key Source Priority

  1. Crossref REST API: DOI and bibliographic metadata validation.
  2. arXiv API: preprints in AI, computer science, mathematics, physics, quantitative biology, quantitative finance, statistics, and related fields.
  3. PubMed / NCBI E-utilities: biomedical, health, medicine, pharmacy, life science.
  4. Europe PMC: biomedical and life science, including open-access signals.
  5. Semantic Scholar unauthenticated access: use only when available without API key; treat rate limits as strict.

Avoid sources that require credentials by default. Do not use Google Scholar scraping.

Agent Contract

Input:

  • topic, research question, chapter title, or claim needing support,
  • discipline,
  • target book type,
  • date range if any,
  • preferred language if any,
  • required source type: journal article, book, conference paper, review, preprint, guideline, or mixed.

Output:

  • status: ready, needs_input, blocked_security, or completed.
  • search_strategy: keywords, synonyms, and no-key sources used.
  • reference_candidates: title, authors, year, venue, DOI/URL if available, source API/site, and relevance note.
  • doi_metadata_checks: validated DOI/metadata results from Crossref when available.
  • open_access_notes.
  • citation_gaps_addressed.
  • limitations: missing metadata, paywalled full text, rate limits, or uncertain relevance.
  • security_privacy_notes.

Process

  1. Check whether the query or source text contains private data or confidential unpublished findings.
  2. Convert the topic into searchable keywords and synonyms.
  3. Choose no-key sources by discipline.
  4. Prefer metadata-rich results with DOI, year, venue, authors, and abstract.
  5. Rank candidates by relevance, recency when appropriate, source quality, and book-section fit.
  6. Execute python .codex/skills/scripts/validate_references.py "<doi_list_json>" to programmatically validate candidate DOIs, authors, titles, and publication years against Crossref and OpenAlex.
  7. Mark anything uncertain instead of inventing details.

Rules

  • Never fabricate titles, authors, journals, DOI, volume, issue, pages, or year.
  • If metadata is incomplete, write [metadata perlu dilengkapi].
  • If a claim still lacks a source, write [perlu sitasi].
  • Do not bypass paywalls or scrape restricted websites.
  • Cache or reuse search results when possible; avoid repeated requests.
  • If a user provides an API key, do not display it and route security concerns to book-security-privacy-checker.