Document Auditor

Problem

Before publishing, teams often spend most of the time asking whether the structure is coherent, the facts are accurate, sensitive details are absent, and the tone is consistent. Manual review is easy to rush, and standards can drift across authors, especially for technical, policy, and public documents.

How it works

doc-auditor checks five fixed dimensions:

• Structure and format: heading levels, lists, code blocks, tables, broken links, missing images
• Content quality: evidence, grammar, logical flow, data freshness
• Factual accuracy: verifiable claims, statistics, dates, terminology
• Security and privacy: personal data, internal paths, internal identifiers, financial data, and other P0 risks
• Style consistency: mixed CJK/Latin text, glossary, tone, punctuation

It classifies the document first, then runs checks in order and reports issues as P0 blockers, P1 serious, P2 moderate, P3 minor. Fast, deep, and change-based modes can cover lightweight, full, or diff-scoped review.

Boundaries

Use it for pre-publish review, change checks, and sensitive-detail screening, not as a legal compliance conclusion or copyright decision. Keep documents under about 50,000 characters. If a file is unreadable, badly formatted, encrypted, or watermarked, clean it up and retry. For high-stakes docs, preserve evidence locations and double-check P0/P1 findings.