Unified citation skill covering discovery, verification, ranking, BibTeX generation, and finalization. Finds citation gaps in LaTeX drafts, retrieves real candidate papers, verifies via academic APIs, ranks with a quality formula, and produces provenance-aware BibTeX. Use when user says "curate citations", "verify references", "find citations", "整理引用", "验证引用", or needs citation management for paper writing.
How this skill is triggered — by the user, by Claude, or both
Slash command
/airesearchorchestrator:citationThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
A unified citation management skill covering the full lifecycle: discovering gaps, searching for papers, verifying authenticity, ranking candidates, generating BibTeX, and producing a finalized citation index.
A unified citation management skill covering the full lifecycle: discovering gaps, searching for papers, verifying authenticity, ranking candidates, generating BibTeX, and producing a finalized citation index.
Scan documents for citation markers that indicate missing references:
[cite], [citation needed], \todo{cite}\cite{} commands against references.bib entries to find uncited references and undefined keysFor each citation gap:
Query academic APIs in priority order:
| Priority | Source | Use |
|---|---|---|
| 1 | Existing project .bib files and verification ledger | Reuse already-verified citations |
| 2 | User-level persistent paper library | Cross-project reuse |
| 3 | Semantic Scholar, OpenAlex | Candidate discovery |
| 4 | CrossRef, DOI resolver | DOI confirmation |
| 5 | DBLP | CS bibliographic records |
API endpoints:
| API | Endpoint |
|---|---|
| Semantic Scholar | api.semanticscholar.org/graph/v1/paper/search |
| CrossRef | api.crossref.org/works |
| arXiv | export.arxiv.org/api/query |
| DBLP | dblp.org/search/publ/api |
| OpenAlex | api.openalex.org/works |
For each candidate, collect: title, authors, year, venue, DOI, abstract, citation count, source URL.
Preprint handling: Treat arXiv/bioRxiv preprints as discovery hints. If a formally published version exists, cite that instead with its DOI. Keep preprints only when no formal version exists, and label them explicitly.
Semantic Scholar API key: Check --semantic-scholar-api-key, then SEMANTIC_SCHOLAR_API_KEY env var, then persisted local secret. Continue with free shared pool if no key is available. Never write API keys into repo files. Never block workflow solely because a key is missing.
For verification sources, grades, and workflow, see references/citation-standards.md (Citation Verification Methodology section).
Verification -- for each citation (existing and new):
Verification by type:
| Citation Type | Required Verification |
|---|---|
| Journal article | DOI verification via CrossRef |
| arXiv preprint | arXiv ID verification |
| Conference paper | DBLP/Semantic Scholar verification |
| Book | ISBN verification |
| Website | URL accessibility check |
Severity levels:
| Issue | Severity |
|---|---|
| Fabricated citation (not in any database) | CRITICAL |
| Wrong claims attributed to source | HIGH |
| Retracted source cited | HIGH |
| Preprint without version number | MEDIUM |
| Incomplete bibliographic info | LOW |
Ranking -- score candidates using the rubric from references/quality-scoring.md:
BibTeX with provenance fields -- for each accepted citation:
@article{author2024method,
title = {Exact Title},
author = {Last, First and Last, First},
year = {2024},
journal = {Venue Name},
doi = {10.xxxx/yyyy},
url = {https://doi.org/10.xxxx/yyyy},
x-bib-source = {crossref},
x-bib-source-url = {https://api.crossref.org/works/10.xxxx/yyyy},
x-verified-with = {semantic-scholar, crossref},
x-verified-at = {2026-03-21},
x-quality-score = {85},
x-verification-status = {verified},
}
Do not fabricate missing BibTeX fields. If the source record is incomplete, note what is missing.
Citation index -- save to paper/citation-index.md with:
Sync and persist: After every run, sync accepted citations into .citation-curator/verification-ledger.json and the user-level paper library. If a target .bib file was provided, sync the ledger with it. Keep user-edited BibTeX entries intact.
See scripts/ for implementation utilities:
| Script | Purpose |
|---|---|
scripts/extract_citation_needs.py | Find citation gaps in .tex files |
scripts/score_papers.py | Rank candidates by quality score |
scripts/fetch_verified_bibtex.py | End-to-end verified citation workflow |
seminal exception.fallback preprint..bib entries -- store extra state in the ledgerreferences/quality-scoring.md - Scoring formula and tie-break rulesreferences/source-verification.md - DOI checks, preprint replacement, BibTeX provenancereferences/citation-standards.md - Citation grading criteria and paper phase citation rulesnpx claudepluginhub jacazjx/ai-research-orchestrator --plugin airesearchorchestratorVerifies academic citations using canonical sources (DOI, arXiv, CrossRef, Semantic Scholar). Provides principles for detecting fake citations and matching metadata.
Plans, audits, and formats academic references. Checks citation authenticity, relevance, retractions, and coverage. Generates BibTeX, GB/T 7714, APA, IEEE, etc.
Searches Google Scholar and PubMed for papers, extracts metadata, validates citations, and generates BibTeX entries. Use when finding papers, converting DOIs, or building bibliographies.