paper-claim-audit

Audit paper .tex claims against raw result evidence files.

Updated May 29, 2026
One-click install
npx skills add https://github.com/Mang30/myskills --skill paper-claim-audit-mang30
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: paper-claim-audit
Source: https://github.com/Mang30/myskills/tree/main/skills/paper-claim-audit
Command: npx skills add https://github.com/Mang30/myskills --skill paper-claim-audit-mang30

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prevents confirmation bias by verifying that every quantitative and scope claim in a paper exactly matches the underlying raw results evidence.

Core Features & Use Cases

  • Zero-context paper-to-evidence audit: Reviews only paper .tex claims and raw result files, with no prior logs or summaries to bias interpretation.
  • Claim-level tracing with verdicts: Extracts each number/comparison/scope claim and maps it to exact evidence values, allowing precise outcomes like PASS/WARN/FAIL.
  • Common integrity failure detection: Flags rounding drift, best-seed cherry-picking, config mismatches, aggregation errors, delta arithmetic mistakes, caption/figure mismatch, and overclaimed evaluation scope.
  • Outputs machine- and human-readable reports: Always writes paper/PAPER_CLAIM_AUDIT.json plus a companion Markdown report for review and downstream automation.

Quick Start

Run the paper-claim-audit skill right after your paper draft is produced and again before submission to ensure every reported number is backed by the raw results.

Frequently Asked Questions about paper-claim-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify that numbers in my research paper match raw experimental results?

To verify numbers in a research paper match raw results, you need a zero-context claim audit that extracts every quantitative statement from .tex files and maps it directly to raw result files. This process outputs a per-claim reconciliation report with PASS, WARN, or FAIL verdicts.

What is zero-context review for paper claim verification?

Zero-context review prevents confirmation bias by auditing paper claims without using prior logs or summaries. The reviewer evaluates only the paper .tex inputs and raw result files independently, ensuring that text interpretation does not drift from actual experimental outputs.

How do I detect best-seed cherry-picking and rounding drift in my manuscript?

Detecting best-seed cherry-picking and rounding drift requires comparing reported paper numbers against raw result files claim by claim. An automated audit flags these common integrity failures alongside config mismatches, aggregation errors, and delta arithmetic mistakes.

Does paper claim auditing work with LaTeX .tex files and raw result evidence?

Yes, paper claim auditing works directly with LaTeX .tex files and raw result evidence files. The process reads only these inputs to extract quantitative and scope claims, mapping them to exact evidence values to produce machine-readable and Markdown audit reports.

When should I run a paper verification audit during the research workflow?

You should run a paper verification audit right after your paper draft is produced and again before final submission. This ensures that every reported number, comparison, and scope claim remains backed by raw results after text edits and improvement loops.

What common integrity failures does a paper claim audit detect?

A paper claim audit detects common integrity failures including rounding drift, best-seed cherry-picking, configuration mismatches, aggregation errors, delta arithmetic mistakes, caption and figure misalignment, and overclaimed evaluation scope by tracing claims to raw evidence.