TheBotique

pubmed-verifier

Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Five-state citation verification (correct / mismatch / partial / invalid / unknown), citation-context parsing, Crossref DOI

as observed 2026-11-20T04:34:16.795Z
Identifier
pubmed-verifier
Source
ClawHub
Version observed
7.2.1
Source repository
not published
Repository observation
No source repository listed
First observed here
2026-09-19T21:21:32.220Z
Observations recorded
38
Installs (reported upstream)
24
Weekly downloads (upstream)
3,559
Declared license
MIT-0

Observation history

2026-11-20T04:34:16.795Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Second evaluation-driven round: the leftover bare module-import group in the entry file is deleted (release notes and source now agree); the package test entry runs a REAL sealed "Privacy hotfix: the package metadata file no longer carries the publisher phone field (the other eight sibling skills never shipped it; platform identity verification lives at the
latestVersion "7.2.0" "7.2.1"
2026-11-19T03:34:19.174Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Evaluation-driven consistency release: the shipped INTEGRITY_MANIFEST no longer lists files the SH whitelist strips (the manifest is rewritten in the same pass); python3 tests/run "Second evaluation-driven round: the leftover bare module-import group in the entry file is deleted (release notes and source now agree); the package test entry runs a REAL sealed
latestVersion "7.1.0" "7.2.0"
2026-11-18T02:34:26.003Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Feedback-driven release: --self-check gains a live retraction-source health probe (both registries must agree on a retracted sample; unreachable registries skip honestly); --title "Evaluation-driven consistency release: the shipped INTEGRITY_MANIFEST no longer lists files the SH whitelist strips (the manifest is rewritten in the same pass); python3 tests/run
latestVersion "7.0.0" "7.1.0"
2026-11-12T21:45:54.906Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "One home per fact: the Top-10 NOT-to-do table is now the speed index atop the Anti-patterns section and Quick FAQ routes there instead of repeating it (EN+ZH, test-locked). README "Feedback-driven release: --self-check gains a live retraction-source health probe (both registries must agree on a retracted sample; unreachable registries skip honestly); --title
latestVersion "6.8.0" "7.0.0"
summary "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv
2026-11-07T16:42:40.757Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "The Files table in the main doc now reads the shipped package: every listed path exists, every package-carriable sample appears by its full name (short forms like .json or +_ZH hi "One home per fact: the Top-10 NOT-to-do table is now the speed index atop the Anti-patterns section and Quick FAQ routes there instead of repeating it (EN+ZH, test-locked). README
latestVersion "6.7.0" "6.8.0"
2026-11-06T15:44:01.875Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "One count, one home per fact: the module count is now entry file + six pv_* modules everywhere (one page still said five while scripts/ holds six - the evaluation quoted the stale "The Files table in the main doc now reads the shipped package: every listed path exists, every package-carriable sample appears by its full name (short forms like .json or +_ZH hi
latestVersion "6.6.0" "6.7.0"
2026-11-05T14:44:09.713Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "placeholder" "One count, one home per fact: the module count is now entry file + six pv_* modules everywhere (one page still said five while scripts/ holds six - the evaluation quoted the stale
latestVersion "6.5.0" "6.6.0"
2026-11-04T13:43:41.465Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "The bibliography samples finally travel: --bibliography, --lint-claims and --claims-template now sniff the format from the file content (BibTeX @type{ and RIS TY - markers), so r "placeholder"
latestVersion "6.3.0" "6.5.0"
2026-11-02T11:41:37.154Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "The consistency gate held (evaluation 4.853, docQuality back to 5.0). This release ships a way to prove a download: --self-check runs five core promises offline (the five-state la "The bibliography samples finally travel: --bibliography, --lint-claims and --claims-template now sniff the format from the file content (BibTeX @type{ and RIS TY - markers), so r
latestVersion "6.2.0" "6.3.0"
2026-10-31T09:43:05.699Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "references/operations.md — the playbook: batching math (per-path API cost table + a worked 3,000-citation example), three real proxy scenarios with --check-net as the arbiter, a s "The consistency gate held (evaluation 4.853, docQuality back to 5.0). This release ships a way to prove a download: --self-check runs five core promises offline (the five-state la
latestVersion "5.5.0" "6.2.0"
2026-10-23T01:44:40.541Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "--help is grouped by task domain: 38 flags reorganized from one flat list into five groups (inputs / verification depth / output & exports / network & performance / institutional "references/operations.md — the playbook: batching math (per-path API cost table + a worked 3,000-citation example), three real proxy scenarios with --check-net as the arbiter, a s
latestVersion "5.4.0" "5.5.0"
2026-10-20T23:50:05.613Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Quantified input guard: --max-entries (default 2000) counts the de-duplicated workload and prints the real number, expected cost and a split recommendation when exceeded — 0 silen "--help is grouped by task domain: 38 flags reorganized from one flat list into five groups (inputs / verification depth / output & exports / network & performance / institutional
latestVersion "5.2.0" "5.4.0"
2026-10-19T22:50:38.661Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Structured retraction detail: every retraction carries retraction_type (retraction vs removal — the latter newly detected via Crossref updated-by and capped the same way), wired t "Quantified input guard: --max-entries (default 2000) counts the de-duplicated workload and prints the real number, expected cost and a split recommendation when exceeded — 0 silen
latestVersion "5.1.0" "5.2.0"
2026-10-18T21:45:32.509Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "pubmed-verifier 5.0.0. 觉得有用的话,欢迎 Star(GitHub)或收藏(SkillHub 技能页)。" "Structured retraction detail: every retraction carries retraction_type (retraction vs removal — the latter newly detected via Crossref updated-by and capped the same way), wired t
latestVersion "5.0.0" "5.1.0"
2026-10-17T20:48:50.018Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "pubmed-verifier 4.9.0. 觉得有用的话,欢迎 Star(GitHub)或收藏(SkillHub 技能页)。" "pubmed-verifier 5.0.0. 觉得有用的话,欢迎 Star(GitHub)或收藏(SkillHub 技能页)。"
latestVersion "4.9.0" "5.0.0"
2026-10-16T19:37:29.109Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "pubmed-verifier 4.7.0. 觉得有用的话,欢迎 Star(GitHub)或收藏(SkillHub 技能页)。" "pubmed-verifier 4.9.0. 觉得有用的话,欢迎 Star(GitHub)或收藏(SkillHub 技能页)。"
latestVersion "4.7.0" "4.9.0"
2026-10-15T18:48:12.129Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Finding ledger closes every report (answered / partially / unresolved with next actions and a budget line); --claims-template reverse-generates a claims CSV skeleton from any refe "pubmed-verifier 4.7.0. 觉得有用的话,欢迎 Star(GitHub)或收藏(SkillHub 技能页)。"
latestVersion "4.6.0" "4.7.0"
2026-10-14T17:39:07.962Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "Submission evidence pack: --report-bundle DIR writes everything a journal or reviewer may ask for in one command — audit.json (replayable working-paper) + report.html + verified-r "Finding ledger closes every report (answered / partially / unresolved with next actions and a budget line); --claims-template reverse-generates a claims CSV skeleton from any refe
latestVersion "4.4.0" "4.6.0"
summary "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv
2026-10-13T16:37:28.889Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Retraction replacement suggestions: with --suggest, a retracted entry now carries up to three same-topic live candidates (retracted papers and its own PMID excluded, marked replac "Submission evidence pack: --report-bundle DIR writes everything a journal or reviewer may ask for in one command — audit.json (replayable working-paper) + report.html + verified-r
latestVersion "4.2.0" "4.4.0"
2026-10-11T14:42:02.570Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "OpenAlex is now the fourth metadata source: a DOI that Crossref reports 404 gets a second look at OpenAlex before any verdict — DataCite/Zenodo/Chinese-registry DOIs become fully "Retraction replacement suggestions: with --suggest, a retracted entry now carries up to three same-topic live candidates (retracted papers and its own PMID excluded, marked replac
latestVersion "4.1.0" "4.2.0"
2026-10-10T14:15:52.413Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "Plain-text reference lists and formatted reference list export. --parse-text refs.txt verifies a reference list copied straight out of a manuscript draft: numbered entries split c "OpenAlex is now the fourth metadata source: a DOI that Crossref reports 404 gets a second look at OpenAlex before any verdict — DataCite/Zenodo/Chinese-registry DOIs become fully
latestVersion "4.0.0" "4.1.0"
summary "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv
2026-10-09T13:20:37.333Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "Chinese references (GB/T 7714) now parse: title/authors/journal/year/DOI extraction from contexts like ……标题[J]. 刊名, 年… PMID: xxx, with CJK-vs-Latin title/journal comparisons skipp "Plain-text reference lists and formatted reference list export. --parse-text refs.txt verifies a reference list copied straight out of a manuscript draft: numbered entries split c
latestVersion "3.9.0" "4.0.0"
summary "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv
2026-10-08T12:19:32.804Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Verdict-ladder corrections from an independent real-data audit: scan-mode author lines are no longer taken as claimed titles (B1 false positives); an entirely different claimed au "Chinese references (GB/T 7714) now parse: title/authors/journal/year/DOI extraction from contexts like ……标题[J]. 刊名, 年… PMID: xxx, with CJK-vs-Latin title/journal comparisons skipp
latestVersion "3.8.0" "3.9.0"
2026-10-07T11:52:53.866Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "RIS bibliography support: --bibliography refs.ris (Zotero/EndNote/Mendeley exports) routes records by PMID (AN tag or note) > DOI (DO) > arXiv (UR/eprint) with full claimed-metada "Verdict-ladder corrections from an independent real-data audit: scan-mode author lines are no longer taken as claimed titles (B1 false positives); an entirely different claimed au
latestVersion "3.7.0" "3.8.0"
2026-10-06T10:50:34.888Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "BibTeX bibliography audit: --bibliography refs.bib verifies a .bib file directly — entries route by PMID (field or note, including notes --export-bibtex itself writes) > DOI > arX "RIS bibliography support: --bibliography refs.ris (Zotero/EndNote/Mendeley exports) routes records by PMID (AN tag or note) > DOI (DO) > arXiv (UR/eprint) with full claimed-metada
latestVersion "3.6.0" "3.7.0"
2026-10-05T10:33:23.015Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "Claims lint: --lint-claims validates a claims file offline before any run (ID shapes, missing titles, unknown/typo columns, duplicates, DOI prefix — zero network, exit-coded). Doc "BibTeX bibliography audit: --bibliography refs.bib verifies a .bib file directly — entries route by PMID (field or note, including notes --export-bibtex itself writes) > DOI > arX
latestVersion "3.5.0" "3.6.0"
summary "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv
2026-10-04T09:46:17.065Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "DOI claims become first-class: claims rows keyed by DOI (no PMID/arXiv) now get the full verdict ladder (correct/mismatch/partial) against the linked PubMed record or Crossref met "Claims lint: --lint-claims validates a claims file offline before any run (ID shapes, missing titles, unknown/typo columns, duplicates, DOI prefix — zero network, exit-coded). Doc
latestVersion "3.4.0" "3.5.0"
2026-10-03T08:46:10.142Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "arXiv preprint-to-published-version cross-check: claimed DOI vs the arXiv-registered version-of-record DOI (agreement = evidence, mismatch caps at partial); bare preprint citation "DOI claims become first-class: claims rows keyed by DOI (no PMID/arXiv) now get the full verdict ladder (correct/mismatch/partial) against the linked PubMed record or Crossref met
latestVersion "3.3.0" "3.4.0"
2026-10-02T07:46:18.975Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "arXiv ID verification + three citation types in one audit. 109 offline + 16 realnet green." "arXiv preprint-to-published-version cross-check: claimed DOI vs the arXiv-registered version-of-record DOI (agreement = evidence, mismatch caps at partial); bare preprint citation
latestVersion "3.2.0" "3.3.0"
2026-10-01T06:47:50.210Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "Milestone: arXiv ID verification — arXiv:2401.12345 and arxiv.org/abs patterns are extracted from scans (or --arxivs) and checked against the official arXiv API; a nonexistent ID "arXiv ID verification + three citation types in one audit. 109 offline + 16 realnet green."
latestVersion "3.0.0" "3.2.0"
summary "Batch-verify PMID citations against PubMed and catch the hallucination that existence checks miss — a REAL PMID pointing to a DIFFERENT paper (the most common AI-fabricated citati "Reference checker for AI-fabricated citations: batch-verify PMIDs against PubMed and catch the hallucination existence checks miss — a REAL PMID pointing to a DIFFERENT paper. Fiv
2026-09-29T04:47:03.526Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "DOI-native verification: --dois verifies DOIs without PMIDs; --source scans auto-extract DOIs from your files; Crossref not-found on an explicitly provided DOI = fabrication signa "Milestone: arXiv ID verification — arXiv:2401.12345 and arxiv.org/abs patterns are extracted from scans (or --arxivs) and checked against the official arXiv API; a nonexistent ID
latestVersion "2.8.0" "3.0.0"
2026-09-28T03:46:57.521Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Retraction detection for EVERY PMID: the registry publication type (Retracted Publication; NCBI + Europe PMC) flags retracted papers with no DOI and no flags needed — and survives "DOI-native verification: --dois verifies DOIs without PMIDs; --source scans auto-extract DOIs from your files; Crossref not-found on an explicitly provided DOI = fabrication signa
latestVersion "2.7.0" "2.8.0"
2026-09-27T02:45:50.156Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Audit working-paper: --export-audit writes a self-contained JSON trail (tool identity, redacted invocation, per-citation evidence chain, verdict trace) a third party can replay. H "Retraction detection for EVERY PMID: the registry publication type (Retracted Publication; NCBI + Europe PMC) flags retracted papers with no DOI and no flags needed — and survives
latestVersion "2.6.0" "2.7.0"
2026-09-26T01:46:07.395Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Author-name verification: single-letter initials never take part in surname matching (no more false author hits); CJK/Latin cross-language author names are honestly skipped, not c "Audit working-paper: --export-audit writes a self-contained JSON trail (tool identity, redacted invocation, per-citation evidence chain, verdict trace) a third party can replay. H
latestVersion "2.5.0" "2.6.0"
2026-09-25T00:46:42.104Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "DOI-splice cross-check: claims now accept a doi field; a claimed DOI differing from the registered one is flagged (capped at partial, exit 1) — no extra API calls. Bidirectional N "Author-name verification: single-letter initials never take part in surname matching (no more false author hits); CJK/Latin cross-language author names are honestly skipped, not c
latestVersion "2.4.0" "2.5.0"
summary "Batch-verify PMID citations against PubMed and catch the hallucination that existence checks miss: a REAL PMID pointing to a DIFFERENT paper (the most common AI-fabricated citatio "Batch-verify PMID citations against PubMed and catch the hallucination that existence checks miss — a REAL PMID pointing to a DIFFERENT paper (the most common AI-fabricated citati
2026-09-24T00:36:14.127Z

Fields that differed: changelog latestVersion

FieldBeforeAfter
changelog "Retraction detection: with --verify-doi, papers Crossref lists as RETRACTED are flagged and their verdict capped at partial with a human-review note (retracted citations exit 1). "DOI-splice cross-check: claims now accept a doi field; a claimed DOI differing from the registered one is flagged (capped at partial, exit 1) — no extra API calls. Bidirectional N
latestVersion "2.3.0" "2.4.0"
2026-09-20T22:20:20.811Z

Fields that differed: changelog latestVersion summary

FieldBeforeAfter
changelog "Network hardening: NCBI API key support (faster batches), automatic Europe PMC fallback when NCBI is unreachable (--meta-source), Crossref polite pool via --mailto, 429 Retry-Afte "Retraction detection: with --verify-doi, papers Crossref lists as RETRACTED are flagged and their verdict capped at partial with a human-review note (retracted citations exit 1).
latestVersion "2.2.0" "2.3.0"
summary "Batch-verify PMID citations against PubMed and catch the hallucination that existence checks miss: a REAL PMID pointing to a DIFFERENT paper (the most common AI-fabricated citatio "Batch-verify PMID citations against PubMed and catch the hallucination that existence checks miss: a REAL PMID pointing to a DIFFERENT paper (the most common AI-fabricated citatio

Correction

If you maintain this extension and believe anything above is inaccurate, request a correction. Corrections are published, and disputed entries are marked as disputed while under review.