Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add chgagne/claude-skills-research --skill verifying-bibliographygit clone --depth 1 https://github.com/chgagne/claude-skills-researchWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/chgagne/claude-skills-research/verifying-bibliography)<a href="https://agentmods.dev/skills/chgagne/claude-skills-research/verifying-bibliography"><img src="https://agentmods.dev/badge/skills/chgagne/claude-skills-research/verifying-bibliography/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/chgagne/claude-skills-research/verifying-bibliography"><img src="https://agentmods.dev/badge/skills/chgagne/claude-skills-research/verifying-bibliography.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00088 | $0.02489 |
| Opus 5 | $0.00044 | $0.01244 |
| Sonnet 5 | $0.00018 | $0.00498 |
| Haiku 4.5 | $0.00009 | $0.00249 |
Grade A, and why
verifying-bibliography scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verifying Bibliography
Overview
Resolve a stable identifier first, then compare the record back against the .bib.
Everything else in this skill follows from that one inversion.
The obvious design — take the title in the .bib, search for it, fuzzy-match the authors —
cannot detect the single most important defect. A fabricated title that accurately describes
the right paper retrieves that paper and passes. Searching by the thing you are trying to
verify is circular.
Two measured examples, both cleared by a title-search tool and both caught by reverse lookup:
moritz2019dracocarried the title "Draco: An Approach to Generating Visualizations with a Computational Design Process". Its DOI is correct and resolves to "Formalizing Visualization Design Knowledge as Constraints: Actionable and Extensible Models in Draco".wongsuphasawat2016voyagercarried DOI10.1109/TVCG.2015.2467251, which resolves to "High-Quality Ultra-Compact Grid Layout of Grouped Networks" — a different paper by six different authors.
Run it
python3 ~/.claude/skills/verifying-bibliography/assets/run-bibcheck.py \
refs.bib --bbl main.bbl --out review-assets/
Run it by absolute path from the paper directory. Stdlib only — no install, no venv,
no dependencies. (python3 -m bibcheck also works, but only from <skill>/assets/.)
--bbl main.bblrestricts checking to entries actually cited (the.bblis ground truth for "cited", because it reflects the built document rather than every\citeincluding commented-out ones)--out DIRwritesbibcheck-report.mdandbibdiff.csv- Exit code
2means source coverage was degraded — see When a source is down
Optional: S2_API_KEY in the environment, or ~/.config/scholarly/s2_key. Semantic Scholar
sits last in the ladder; on two real bibliographies it changed nothing. Absent, it is
silently skipped and nothing breaks.
The ladder
1. DOI -> Crossref /works/{doi} authoritative, reverse check
2. arXiv DOI -> arXiv API 10.48550/* is DataCite, absent from Crossref
3. arXiv id -> arXiv API
4. ACL DOI -> aclanthology.org via Crossref
5. no identifier -> DBLP by title the CS/ML spine
6. still nothing -> OpenAlex, then Semantic Scholar
What ships with it
20 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- assets/bibcheck/__init__.py 0 B runs code
- assets/bibcheck/__main__.py 2.3 KB runs code
- assets/bibcheck/bibparse.py 304 B runs code
- assets/bibcheck/compare.py 9.9 KB runs code
- assets/bibcheck/normalize.py 417 B runs code
- assets/bibcheck/report.py 2.8 KB runs code
- assets/bibcheck/sources.py 621 B runs code
- assets/run-bibcheck.py 591 B runs code
- assets/tests/test_bibparse.py 2.0 KB runs code
- assets/tests/test_breaker.py 4.3 KB runs code
- assets/tests/test_compare.py 14 KB runs code
- assets/tests/test_normalize.py 6.1 KB runs code
- assets/tests/test_preprint_policy.py 5.3 KB runs code
- assets/tests/test_report.py 2.4 KB runs code
- assets/tests/test_retry.py 1.9 KB runs code
- assets/tests/test_s2_key.py 1.9 KB runs code
- assets/tests/test_sources_arxiv_doi.py 710 B runs code
- assets/tests/test_sources_offline.py 2.1 KB runs code
- assets/tests/test_stdlib_only.py 982 B runs code
- reference/fabrication-signatures.md 8.2 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 190 lines · 88 tokens per session scan A dc069adfdd4d
verifying-bibliography is a skill published in the GitHub repository chgagne/claude-skills-research (4 stars, last pushed 7d ago), licensed MIT. It adds 88 tokens to every session and 2,489 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
literaturverzeichnispruefung
Verify academic references and bibliographies for citation errors and AI hallucinations using live database APIs (OpenAlex, OpenAIRE, K10plus, lobid, DNB, Crossref, Semantic Scholar, ZBW EconBiz, Open Library, arXiv, Library of Congress, GESIS). Trigger whenever a user wants to check, verify, or validate ANY…
secure-auth
Secure authentication patterns (OWASP, NIST). Use for login, registration, password reset, sessions, JWT, OAuth, MFA, passkeys.
academic-writing
Scholarly writing and research compliance. Use for CRediT, preregistration, Plan S, Nelson Memo, preprints, ORCID, LLM disclosure.
page-monitoring
Web page change detection, availability tracking, and RSS feed generation. Use to monitor changes, downtime, or make a feed.
web-archiving
Web archiving and retrieval via Wayback Machine and Archive.today. Use to preserve content, reach dead pages, or save evidence.
accessibility-compliance
Web accessibility patterns for news and academic sites. Use for WCAG audits, alt text, accessible data viz, and assistive tech.