Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

Integrity Auditor

Audits a paper (PDF/DOI/arXiv) across image, numerical, and logical evidence tracks and produces a Level 1–4 graded, reviewable evidence report.

Data & AnalyticsAdvanced★ 185⑂ 17AI score 9/10Last updated: Jul 28, 2026

What it does

  • Detects the input mode: local PDF, DOI/arXiv ID, or a local paper-writer run slug.
  • Works around paywalls by harvesting publisher CDN endpoints for supplementary files (MOESM XLSX), hi-res figures, and Author Correction PDFs — including the correction's own supplement, which often holds pre-correction figure originals.
  • Triages the paper class first (bio / ML / theory-review) so only the applicable checks run.
  • Runs three evidence tracks:
    • Image: panel splitting → perceptual-hash duplicate detection → ORB feature matching for rotated/flipped/cropped duplicates, Western-blot background continuity, fluorescence channel label swaps.
    • Numerical: n-consistency, mean/SD/SEM recomputation, P-value sanity, last-decimal pattern sweeps, unit/order-of-magnitude mismatches, cross-XLSX aggregate disagreements, and for ML papers arithmetic re-derivation of every quoted improvement.
    • Logical: compress claims into "A through B causes C", check for missing controls and replication gaps.
  • Grades every finding Level 1–4 with reviewable pointers (figure id, page/line, transformation description, and the exact raw data to request from authors).
  • Never issues verdicts. It produces evidence and grading only, plus a "Requested raw data" section.

Who it's for

  • Peer reviewers, journal editors, and research-integrity officers.
  • PhD students and PIs self-auditing a manuscript before submission.
  • Teams using AI-assisted paper drafting who must reconcile every reported number against experiment results.json.

Examples

  1. "Check whether Figure 3 panels in this DOI are duplicated" → pulls hi-res figures and supplements from the CDN, splits panels, runs phash then escalates to ORB, and writes graded finding files.
  2. "Do these two Nature supplementary tables report the same total differently?" → xlsx_aggregate_consistency.py compares sums and logs a 0.62% systematic gap as a Level 1 finding.
  3. "Verify the '+1.2pp improvement' claim in this arXiv ML paper" → image track records _inapplicable.md; the numerical track re-derives every delta from the benchmark cells and checks seed/std reporting consistency (Check 5).

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/ai4s-research/ai4s-skills/HEAD/skills/integrity-auditor/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/integrity-auditor folder from the GitHub repo ai4s-research/ai4s-skills into my ~/.claude/skills/integrity-auditor/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/ai4s-research/ai4s-skills.git && mkdir -p ~/.claude/skills && cp -r ai4s-skills/skills/integrity-auditor ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal (on Windows, WSL is recommended).
  2. Install prerequisites: macOS brew install poppler; Ubuntu sudo apt install poppler-utils. For the image/XLSX tools also run pip install pillow opencv-python numpy openpyxl.
  3. Clone the repo: git clone https://github.com/ai4s-research/ai4s-skills.git
  4. Copy the skill into place: mkdir -p ~/.claude/skills && cp -r ai4s-skills/skills/integrity-auditor ~/.claude/skills/
  5. Verify with ls ~/.claude/skills/integrity-auditor that SKILL.md, references/, templates/, and forensics_tools/ exist (the doc contains a stale line saying v1 ships without forensics_tools/, so confirm).
  6. Restart Claude Code, then ask: "Audit this paper for integrity issues" with a PDF path, DOI, or arXiv ID.
  7. Output lands in output/integrity-auditor/<slug>/latest/audit_report.md under your working directory.
View source on GitHub ↗License: MIT