Works with the AI Agents you already use
DOCX Redline & Tracked Changes Workbench
Create, inspect, and redline .docx files with real tracked changes, comments, and formatting fidelity — auditable Word reviews that never corrupt the source.
Free
DOCX Redline & Tracked Changes Workbench
Example session with this skill installed
Review vendor_msa_draft_v3.docx for our counsel and redline it: change the cure period in Section 4.2 from 30 days to 60 days, add a liability cap of 12 months of fees in Article 9, and flag the indemnity language in 7.3 with a comment. Leave every edit as a tracked change authored 'Edric Review' with proper w:del and w:ins markup — never silently replace the source text. After packing, verify with pandoc --track-changes=all that each edit appears, numbering and tables still render, and no duplicate revision IDs exist. Give me the reviewed docx plus a change summary.
- Read your context and instructions
- Compiled the docx redline
Summary
Redlined vendor_msa_draft_v3.docx to vendor_msa_draft_v3_reviewed.docx: 6 tracked insertions, 4 deletions, 1 comment, all authored "Edric Review" with unique w:id values. The original file is untouched and every change is auditable in Word's review pane.
| Location | Original | Tracked change | Verification |
|---|---|---|---|
| §4.2 cure period | 30 days | 60 days (w:del + w:ins) | pandoc --track-changes=all: present |
| Art. 9 liability | no cap text | Cap = 12 months of fees | new runs, rPr cloned |
| §7.3 indemnity | as-is | comment: scope unclear | comments.xml rel linked |
| §1 definitions | "Vendor" | "Contractor" (2 runs patched) | split runs re-joined correctly |
Final checks: w:id 501–512 unique, headings/numbering intact, tables not collapsed, sectPr unchanged, PDF conversion clean with correct comment pane.
Next steps
- Open in Word and confirm the 11 revisions render in the review pane
- Route §7.3 comment to counsel for accept/reject
- On sign-off, publish as MSA-v4 and archive the working copy
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Standard LLMs often struggle with Word documents, either stripping critical formatting or corrupting the underlying OOXML structure. Developers need a way to programmatically edit .docx files while preserving tracked changes, comments, and complex styles like legal numbering.
What it does
- Extracts plain text, headings, and tables while preserving or accepting tracked changes using Pandoc.
- Manages Word document revision history by inserting valid
w:insandw:delOOXML tags. - Handles complex document parts including comments, relationships, and style definitions without corrupting the ZIP package.
- Automates visual quality assurance by converting documents to PDF and images for layout verification.
Frameworks & tools
Pandoc, LibreOffice, Node.js docx library, Python lxml, and Poppler-utils.
Why this beats prompting it yourself
Simple prompting usually results in plain text output that loses all Word-specific metadata. This skill provides the exact XML structures and workflow patterns needed to manipulate OOXML directly, ensuring that "Tracked Changes" actually show up in the final Microsoft Word file.
Use cases
- Automating legal redlines and contract reviews with auditable tracked changes.
- Generating branded corporate memos and reports using Node.js or Python.
- Programmatically inserting comments and annotations into existing academic or government documents.
- Batch verifying document layout and formatting via automated visual snapshots.
Known limitations
Requires environment-level tools like Pandoc or LibreOffice for full extraction and PDF verification. Direct XML search is brittle if Word splits text across multiple runs.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse All Claude Skills.
Targets OOXML .docx only — legacy .doc, RTF, and ODF need conversion first. Heavy structural edits (section rebuilds, numbering recreation) still benefit from a manual pass in Word.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
2 uses
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 2 months ago
- Free with an Agensi account
- Software
Python
Permissions
Allowed Hosts
File Scopes
Terminal plus read/write on the working copy of the .docx package; optional registry.npmjs.org access only to install the docx library for brand-new document generation.