sci-audit

v0.2.5

Domain-agnostic forensic and linguistic auditor for LLM-generated scientific text. Seven audit axes: reference integrity, claim grounding (with legal open-access full-text resolution), document-internal statistical consistency (statcheck/GRIM/GRIMMER/SPRITE/CI/percentage/subgroup, Turkish decimals), hallucination signals (with ISBN/ORCID/arXiv checksums + entity verification), reporting-guideline conformance (14 guidelines + section pre-scan), AI-use transparency (ICMJE/COPE), and Turkish scientific writing/orthography. Deterministic stdlib core with graceful provider degrade; remote-only MCP; Claude-native judge; injection shield + privacy invariant + content-hashed evidence ledger.

CodexClaude Code7 Skills4 MCP serversStreamable HTTP

By CureonicsLicense: MIT0 GitHub starsUpdated 1 hour ago

Directory evidence

Runtimes: Codex and Claude Code
Parsed components: 11 skill or MCP entries
Source updated: Oct 2, 2026
Manifest status: Canonical path parsed

The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →

Install sci-audit for Codex and Claude Code

Installs for the current user

codex plugin marketplace add mahirkurt/CureoPrivate
codex plugin marketplace upgrade cureonics-private
codex plugin add sci-audit@cureonics-private

Paste and run these commands in a terminal with Codex. They add and refresh the cureonics-private catalog, then install this plugin.

Compatibility: the page URL and API slug “sci-audit” remain stable.

Codex: sci-audit@agent-plugin-marketplace → sci-audit@cureonics-private

The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.

Get the source manually

$ git clone https://github.com/mahirkurt/CureoPrivate

Clone the source repository, then follow its setup instructions to add the plugin to a compatible client. The plugin root is plugins/sci-audit/.

Plugin files

plugins/sci-audit/

├── .codex-plugin/plugin.json
├── .claude-plugin/plugin.json
├── skills/ai-transparency/SKILL.md
├── skills/citation-forensics/SKILL.md
├── skills/claim-grounding/SKILL.md
├── skills/hallucination-signals/SKILL.md
├── skills/sci-audit-orchestrator/SKILL.md
├── skills/stats-forensics/SKILL.md
├── skills/turkish-sci-style/SKILL.md
└── .mcp.json

Included Skills7

ai-transparencyskills/ai-transparency/SKILL.md

Use when checking whether an LLM-generated scientific text discloses its AI use (axis F) — trigger on "is AI use disclosed", "check AI transparency", "ICMJE AI statement", "yapay zeka beyanı var mı", "AI disclosure", "undisclosed AI". Scans for a present/absent AI-use disclosure statement (ICMJE/COPE/WAME) and LLM giveaway signals, and flags unfilled placeholders.

citation-forensicsskills/citation-forensics/SKILL.md

Use when auditing the references of an LLM-generated scientific text for fabricated or hallucinated sources (axis A) — trigger on "are these citations real", "check the DOIs", "verify the references", "did the model make up this paper", "kaynak uydurma kontrolü". Resolves each DOI/PMID/arXiv id against remote scholarly MCPs, matches title/author/year/journal metadata, and checks retraction status.

claim-groundingskills/claim-grounding/SKILL.md

Use when checking whether the numeric and factual claims in an LLM-generated scientific text are actually supported by a cited source (axis B) — trigger on "is this number sourced", "check claim grounding", "does the citation support this", "kaynaksız iddia var mı", "iddia temellendirme". Extracts claims, checks each for a nearby source marker, and escalates to evidence matching against scholarly MCPs.

hallucination-signalsskills/hallucination-signals/SKILL.md

Use when scanning an LLM-generated scientific text for hallucination and confabulation signals (axis D) — trigger on "does this sound made up", "check for hallucination", "over-certainty language", "invented method names", "halüsinasyon sinyalleri", "aşırı kesinlik". Flags over-certainty absolutes, universal quantifiers, malformed identifiers, and uncited named methods; optionally runs semantic entropy with a sampling backend.

sci-audit-orchestratorskills/sci-audit-orchestrator/SKILL.md

Use when auditing an LLM-generated scientific text (article draft, thesis chapter, review, report, abstract) for integrity and language quality — trigger on "audit this text", "check this draft", "is this citation real", "verify these stats", "bilimsel metni denetle", "kaynakları doğrula", "Türkçe imla kontrol". Detects language (Turkish auto-enables axis G), section type, splits long text, fans out the seven audit axes, and merges the scores into one report.

stats-forensicsskills/stats-forensics/SKILL.md

Use when checking the internal statistical consistency of an LLM-generated scientific text (axis C) — trigger on "check the stats", "do these p-values add up", "statcheck this", "GRIM test", "are these percentages consistent", "istatistik tutarlılığı", "p değeri kontrolü". Recomputes p-values from reported test statistics, runs GRIM on reported means, and flags impossible effect sizes.

turkish-sci-styleskills/turkish-sci-style/SKILL.md

Use when auditing Turkish scientific text for writing quality and orthography (axis G) — auto-triggers when the text is Turkish or the user passes --lang tr. Trigger on "Türkçe imla denetimi", "bilimsel Türkçe kontrol", "ondalık virgül", "akademik dil denetimi", "check Turkish scientific writing", "APA-TR sayı biçimi". Runs the deterministic G1–G6 core (tr_sciaudit.py), optional TDK/Zemberek/GECTurk providers with safe degrade, and the style-judge subagent for register/fluency.

MCP servers4

openalexStreamable HTTP

url: https://openalex.cureonics.com/mcp

pubmedStreamable HTTP

url: https://pubmed.mcp.claude.com/mcp

pubmed-epmcStreamable HTTP

url: https://pubmed.cureonics.com/mcp

semantic-scholarStreamable HTTP

url: https://semanticscholar.cureonics.com/mcp

Plugin manifests2

plugins/sci-audit/.codex-plugin/plugin.json

{
  "name": "sci-audit",
  "version": "0.2.5",
  "description": "Domain-agnostic forensic and linguistic auditor for LLM-generated scientific text. Seven audit axes: reference integrity, claim grounding (with legal open-access full-text resolution), document-internal statistical consistency (statcheck/GRIM/GRIMMER/SPRITE/CI/percentage/subgroup, Turkish decimals), hallucination signals (with ISBN/ORCID/arXiv checksums + entity verification), reporting-guideline conformance (14 guidelines + section pre-scan), AI-use transparency (ICMJE/COPE), and Turkish scientific writing/orthography. Deterministic stdlib core with graceful provider degrade; remote-only MCP; Claude-native judge; injection shield + privacy invariant + content-hashed evidence ledger.",
  "author": {
    "name": "Cureonics",
    "url": "https://cureonics.com"
  },
  "homepage": "https://cureonics.com",
  "license": "MIT",
  "keywords": [
    "scientific-integrity",
    "hallucination",
    "citations",
    "turkish",
    "imla",
    "guardrails",
    "reporting-guidelines",
    "statcheck",
    "claim-grounding",
    "ai-transparency"
  ],
  "skills": "./skills/",
  "interface": {
    "displayName": "Scientific Audit",
    "shortDescription": "Bilimsel metinler için atıf, iddia, istatistik ve Türkçe bilim dili denetimi.",
    "longDescription": "LLM tarafından üretilmiş bilimsel metinleri hem adli hem dilsel denetleyen genel amaçlı, alan-bağımsız araç. Yedi eksen: (A) referans bütünlüğü, (B) iddia temellendirme + yasal açık-erişim TAM METİN doğrulama, (C) belge-içi istatistik tutarlılığı (statcheck/GRIM/GRIMMER/SPRITE/GA/yüzde/altgrup — ham veri gerektirmez, Türkçe ondalık virgül destekli; kaldırılan replicatio'nun sayısal-tutarlılık katmanını stdlib olarak konsolide eder), (D) halüsinasyon sinyalleri + ISBN/ORCID/arXiv checksum + varlık (ilaç/gen/hastalık) doğrulama, (E) 14 raporlama kılavuzu (PRISMA/PRISMA-ScR/CONSORT/CONSORT-AI/STROBE/COREQ/SRQR/JARS/TRIPOD/TRIPOD+AI/STARD/SPIRIT/CHEERS/ARRIVE) + bölüm ön-tarama, (F) AI-kullanım şeffaflığı (ICMJE/COPE), (G) Türkçe bilimsel dil ve imla (otomatik). Deterministik çekirdek SAF Python stdlib (claude.ai web dahil); uzak-yalnız MCP; Claude-native yargıç; injection kalkanı + gizlilik invariantı + content-hash kanıt defteri. No-fabrication: çözülemeyen kaynak 'unverified', asla 'geçti' olmaz.",
    "developerName": "Cureonics",
    "category": "Scientific Integrity",
    "capabilities": [
      "Interactive",
      "Read",
      "Write"
    ],
    "websiteURL": "https://cureonics.com",
    "brandColor": "#B91C1C",
    "defaultPrompt": [
      "Audit this scientific report for citations and claims.",
      "Check statistical consistency in this manuscript.",
      "Review Turkish scientific style and AI transparency."
    ],
    "screenshots": []
  },
  "mcpServers": {
    "pubmed": {
      "type": "http",
      "url": "https://pubmed.mcp.claude.com/mcp",
      "_trust": "Anthropic resmî PubMed MCP (keyless). sci-audit eksen A/B için hesap-seviyesi PubMed; pubmed-epmc (EPMC + Unpaywall tam metin) bunu tamamlar."
    },
    "pubmed-epmc": {
      "type": "http",
      "url": "https://pubmed.cureonics.com/mcp",
      "headers": {
        "Authorization": "Bearer ${PUBMED_MCP_API_KEY}"
      },
      "_trust": "operatör self-host (HP systemd; Evidentia ile aynı uç). Eski Worker DO kotası 1101 verdi (2026-08-17); Hub SQLite cache + 11 araç. KAPILI.",
      "_probe": "2026-08-17 Hub SSOT URL; pubmed_search_articles authors=string[]."
    },
    "openalex": {
      "type": "http",
      "url": "https://openalex.cureonics.com/mcp",
      "headers": {
        "Authorization": "Bearer ${OPENALEX_MCP_API_KEY}"
      },
      "_trust": "operatör self-host (HP systemd; Evidentia ile aynı uç). Eski Worker DO kotası 1101 verdi (2026-08-17); Hub SQLite cache + 5 araç. KAPILI.",
      "_probe": "2026-08-17 Hub SSOT URL; openalex_resolve_name Oxford → total=1."
    },
    "semantic-scholar": {
      "type": "http",
      "url": "https://semanticscholar.cureonics.com/mcp",
      "headers": {
        "Authorization": "Bearer ${SEMANTICSCHOLAR_MCP_API_KEY}"
      },
      "_trust": "operatör self-host (HP systemd; Evidentia ile aynı uç). Eski Worker DO kotası 1101 verdi (2026-08-17); Hub SQLite cache + 4 araç. KAPILI.",
      "_probe": "2026-08-17 Hub SSOT URL; search_papers authors=array of objects."
    }
  }
}

plugins/sci-audit/.claude-plugin/plugin.json

{
  "name": "sci-audit",
  "displayName": "Scientific Audit",
  "version": "0.2.5",
  "description": "Bilimsel metinler için alan bağımsız adli ve dilsel denetçi. Atıf bütünlüğü, iddiaların kaynaklarla doğrulanması, belge içi istatistik tutarlılığı, halüsinasyon sinyalleri, raporlama kılavuzları, AI kullanım şeffaflığı ve Türkçe bilimsel dili denetler. Deterministik Python hesaplamaları ve kanıt defteri kullanır; doğrulanamayan kaynakları unverified olarak bildirir.",
  "author": {
    "name": "Cureonics",
    "url": "https://cureonics.com"
  },
  "homepage": "https://cureonics.com",
  "license": "MIT",
  "keywords": [
    "scientific-integrity",
    "hallucination",
    "citations",
    "turkish",
    "imla",
    "guardrails",
    "reporting-guidelines",
    "statcheck",
    "claim-grounding",
    "ai-transparency"
  ]
}

For maintainers

If you maintain this plugin, link to this source-backed listing from your README so users can review its manifest and indexed components.

[sci-audit on Agent Plugins Marketplace](https://pluginsmp.com/plugins/sci-audit)