An offline document QA skill that finds internal consistency errors in long DOCX, Markdown, and TXT files and produces a repeatable re-check command.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install doc-consistency
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install doc-consistency using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is doc-consistency?
doc-consistency is an Openclaw Skills document QA utility for checking whether a long document is consistent with itself. It detects dangling cross-references, numbering gaps and duplicates, table-of-contents mismatches, missing recurring sections, terminology drift, duplicate headings, stale years, leftover placeholders, and paragraphs that may have been lost during a merge.
The skill runs fully offline with zero runtime dependencies beyond Node.js 18+, so documents remain on the local machine. It does not write content, polish prose, fact-check claims, provide legal or technical judgments, or make semantic decisions; instead, it performs deterministic, located checks and returns a findings summary, detailed issues, and a command that can be rerun for acceptance or CI validation.
doc-consistency Use Cases
- Proofread a textbook, manuscript, thesis, manual, contract set, bid, or RFP response for internal consistency.
- Audit figure, table, equation, appendix, and section cross-references before publication.
- Detect numbering gaps, duplicate labels, unusual starting numbers, and repeated headings.
- Compare a merged document with baseline chapters or source files to identify potentially lost paragraphs.
- Verify that the table of contents agrees with the document body.
- Enforce consistent terminology with a glossary such as
AI Agent = AI agent, AI-Agent. - Find unresolved placeholders including
TODO,TBD,XXX,待填, andLorem ipsum. - Catch suspicious stale years near project, date, or contract references.
- Gate proposal or manuscript delivery when high-severity findings remain.
- Use Openclaw Skills for a repeatable offline document consistency check instead of relying only on manual review.
How doc-consistency Works
- Accept one file or a directory containing supported
.docx,.md, or.txtdocuments. - Extract readable document text and identify headings, section structure, numbering patterns, references, terminology, dates, placeholders, and recurring section signals.
- Compare references against available figures, tables, equations, appendices, chapters, headings, and table-of-contents entries.
- Detect numbering duplicates, gaps, abnormal starts, duplicate same-level headings, missing standard sections, terminology spelling variants, placeholders, and possible stale years.
- When
--baselineis supplied, compare the current document with the baseline paths and flag content that appears to have disappeared during merging. - Group findings by severity and rule type, provide locations and evidence, and show a complete summary even when the displayed list is limited.
- Emit a re-check command so the same acceptance criteria can be rerun manually, in CI, or as part of a contract delivery workflow.
- Clearly report unsupported Word fields such as automatic numbering and automatic footnotes when they are not present in the text layer; an absent finding does not prove those fields are correct.
doc-consistency Setup
Install or access Node.js 18 or newer, then run the checker from the skill directory. The utility has zero third-party dependencies and does not require network access, uploads, or model calls.
node --version
node scripts/check.mjs <file-or-directory> [options]
Common commands:
# Check one DOCX file
node scripts/check.mjs final-manuscript.docx
# Check every supported document in a directory and show all findings
node scripts/check.mjs ./manuscript/ --all
# Compare a merged document with baseline source files
node scripts/check.mjs merged.docx --baseline ./chapters/
# Fail with exit code 1 when high-severity findings exist
node scripts/check.mjs proposal.docx --fail-on high
# Produce complete machine-readable JSON
node scripts/check.mjs manuscript.md --json
# Use a terminology glossary and English reporting
node scripts/check.mjs manuscript.md --glossary glossary.txt --lang en
Available options:
--all: Display all findings instead of the default first 20; the summary remains complete.--limit <n>: Set the maximum number of findings shown in the list.--json: Emit the complete, untruncated JSON result for automation.--glossary <file>: Load one rule per line in the formPreferred = variant1, variant2.--baseline <path...>: Compare against baseline files or directories for merge-loss detection.--fail-on high|medium|low: Exit with code 1 at the selected severity threshold.--lang zh|en: Select the report language.
doc-consistency Data Schema & Taxonomy
The checker is primarily a reporting tool: it reads source documents and emits findings rather than modifying the originals or creating replacement files.
| Data element | Description |
|---|---|
| Source inputs | One file or directory containing .docx, .md, and .txt documents. |
| Location | The document position associated with a finding, such as a heading, paragraph, or referenced item. |
| Rule type | Cross-reference, numbering, TOC/body mismatch, missing section, terminology, placeholder, stale year, duplicate heading, or merge-loss detection. |
| Severity | High, medium, or low, used for prioritization and --fail-on automation. |
| Evidence | The text or structural signal that triggered the finding, such as a missing figure number or inconsistent term. |
| Summary | Counts by severity and distribution by finding type; summaries remain complete even when list output is limited. |
| Re-check command | A rerunnable command included in the report for verification, acceptance, CI, or contract workflows. |
| Glossary metadata | Optional preferred terms and comma-separated variants, one mapping per line. |
| Baseline metadata | Optional reference paths used to identify paragraphs present in baseline material but absent from the current document. |
No network payload, model request, or document upload is required. By default, findings are emitted as command output; --json provides a complete machine-readable result for downstream processing.
doc-consistency Advanced Features
- Fully offline execution with zero dependencies beyond Node.js 18+, supporting privacy-sensitive manuscripts, proposals, and contracts.
- Machine-readable output through
--jsonfor scripts, dashboards, and downstream Openclaw Skills workflows. - Severity-aware quality gates with
--fail-on high|medium|low, suitable for CI pipelines and delivery acceptance. - Baseline comparison for detecting paragraphs or sections lost during multi-author merges.
- Custom terminology governance through a simple glossary file.
- Directory-level scanning for chapter collections and multi-contributor document projects.
- Bilingual reporting with
--lang zhor--lang en. - Controlled output volume using
--alland--limitwhile preserving complete aggregate counts. - Located findings and a reproducible re-check command, making review results auditable rather than dependent on a one-time manual pass.
- Explicit boundary reporting for Word automatic numbering and footnote fields that are unavailable in the text layer, avoiding false assurances.
SKILL.md
Loading
Related Openclaw Skills
Gitcrawl
Gitcrawl provides a freshness-aware GitHub archive workflow for finding issues, triaging pull requests, and identifying duplicate discussion clusters safely.
Frontend UI Engineer
Frontend UI Engineer automates production-quality, accessible, responsive UI development and related engineering workflows with structured outputs and built-in error recovery.
Publish Webpage
Publish Webpage turns a user's idea into a polished, shareable static website and deploys it to goclawgo without requiring code or hosting knowledge.
idea-refine
idea-refine transforms raw, ambiguous ideas into structured, actionable concepts through automated analysis, assumption testing, and refinement.
A-Share Financial Literacy Q&A
An AI-powered A-share financial literacy assistant that explains market concepts, personalizes learning, analyzes risk, and supports simulation-based investment research.
Tri-Platform Video Analysis
Convert a Douyin, Bilibili, or Xiaohongshu video link into a locally transcribed, AI-analyzed, interactive HTML report.
Featured*
