Stats
Actions
Available In
Tags
Plugin
citation-double-check
Verify academic citations: parse references, fetch evidence, check claims against sources, suggest repairs.
README
Citation Double-Check
A Claude Code skill that automatically vets LaTeX paper citations before submission or camera-ready. It catches two categories of citation problems that hurt papers most:
- Hard errors (Track A) — the bib entry itself is wrong: undefined keys, unresolved references, metadata mismatches, missing fields, venue errors, duplicates.
- Soft errors (Track B) — the bib entry is real but the citation is misplaced: the surrounding sentence's claim does not match the cited paper's abstract.
Quick Start
# From a Claude Code session with a LaTeX paper directory:
> check my citations in ~/papers/my-paper
Or trigger with phrases like: "check citations", "vet the bib", "verify references", "camera-ready bib check".
Pipeline
| Stage | What it does | Network? | LLM? |
|---|---|---|---|
| Stage 0a | Parse .tex and .bib files | No | No |
| Stage 0b | Fetch evidence from 8 academic sources | Yes | No |
| Stage 1 | Diagnose unresolved entries | Yes | Yes |
| Track A | Deterministic bib-health checks (A1-A9) | No | No |
| Track B | Claim-vs-abstract semantic checks | No | Yes |
| Stage 4 | Generate report.md + report.json | No | No |
Track A Checks (A1-A9)
| Code | Severity | What it checks |
|---|---|---|
| A1_undefined_key | Critical | \cite{X} where X is missing from bib |
| A2_unused_entry | Cleanup | Bib entry never cited |
| A3_duplicate_entry | Warning | Two keys with fuzzy title similarity ≥ 0.95, same year |
| A4_unresolved | Critical/Warning | All sources missed (downgraded for non-traditional sources) |
| A5_metadata_mismatch | Warning | Title/year/first-author mismatch vs external sources |
| A6_field_missing | Cleanup | Required or recommended BibTeX fields absent |
| A7_venue_mismatch | Cleanup | Bib venue differs from DBLP venue |
| A8_venue_style | Cleanup | Same venue but different formatting |
| A9_bib_entry_error | Critical | Stage 1 detected arXiv ID/DOI/title errors |
Track B Verdicts
| Verdict | Severity | Meaning |
|---|---|---|
| STRONG_SUPPORT | — | Abstract directly supports the claim (not reported) |
| WEAK_SUPPORT | Cleanup | Topically related but abstract lacks specifics |
| OFF_TOPIC | Critical | Abstract topic is completely unrelated |
| WRONG_DIRECTION | Critical | Abstract contradicts the claim |
| INSUFFICIENT_DATA | Warning | Abstract missing or too short |
Output
All output goes to <paper-dir>/runs/citation-check/:
report.md— human-readable, severity-grouped (Chinese descriptions)report.json— machine-readable flat list- Intermediate artifacts:
bib_entries.json,citations.json,evidence_pack.json,track_a_findings.json,track_b_findings.json
Configuration
- Semantic Scholar API key: set
SEMANTIC_SCHOLAR_API_KEYenv var (optional, increases rate limit from 1/s to 100/s) - NCBI API key: set
NCBI_API_KEYenv var (optional, for PubMed)
Dependencies
pip install bibtexparser requests
Running Tests
python3 -m unittest discover -s tests -v
License
MIT
Stats
Version1.0.0
LanguagePython
Stars1
MaintenanceExcellent
LicenseMIT
Last Commit
Added
Available In
Tags
Categories
