Works with the AI Agents you already use
doc privacy
Audit and sanitize Office and PDF files for hidden metadata, tracked changes, and invisible fingerprinting characters.
Free
doc privacy
Example session with this skill installed
Check internal_proposal.docx for any hidden metadata or tracked changes before I send it to the client.
- Read your context and instructions
- Compiled the doc privacy
- Generated the document
I have analyzed internal_proposal.docx by inspecting the document's internal XML structure.
Document Privacy Audit: internal_proposal.docx
| Risk Category | Finding | Details |
|---|---|---|
| Identity Leaks | High | Found dc:creator: 'John Doe' and Company: 'Acme Corp Internal'. |
| Tracked Changes | Medium | 12 tracked insertions and 8 deletions found. Deleted text is recoverable. |
| Hidden Content | Medium | 3 comments found in word/comments.xml. |
| Invisible Characters | Clean | No zero-width spaces or joiners detected. |
| Visual Deception | Low | 2 instances of w:vanish (hidden text) found in the footer. |
| System Info | Info | Created: 2023-10-12, Total Edit Time: 440 minutes, App: Microsoft Word 16.0. |
Privacy Risk Summary
The file contains your full name and company identity. More importantly, it contains
20 tracked changes. Anyone with the file can right-click to "Show Changes" and see your deleted drafts and internal notes.
Next steps
- Open the file in Word, go to the Review tab, and select Accept All Changes and Stop Tracking.
- Delete all comments manually to ensure document integrity.
- Ask me to
clean internal_proposal.docxto wipe the Author and Company metadata fields.
internal_proposal.docx
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Standard office documents often leak more than just their visible text. Sending a file can unintentionally expose author names, internal company paths, deleted drafts, or invisible watermarks. This skill audits .docx, .xlsx, .pptx, and .pdf files to identify privacy risks before they are shared.
What it does
- Metadata extraction identifies author names, last editors, company names, and template file paths.
- Hidden content detection flags tracked changes, comments, hidden Excel sheets, and PowerPoint speaker notes.
- Visual deception audit finds text formatted as white-on-white, 1-point font sizes, or hidden text runs.
- Invisible character scan locates zero-width spaces, joiners, and other non-printing characters used for fingerprinting.
- Privacy cleaning generates a sanitized copy by wiping specific identity fields and reset session timers.
How it works
- Analyze the target file by inspecting the raw XML structure of the zip archive or PDF metadata strings.
- Evaluate risks across four categories: identity leaks, hidden content, invisible characters, and session data.
- Report findings in order of severity, quoting surrounding text for any suspicious invisible characters found.
- Clean the file upon request by creating a new version with empty identity tags and reset revision counts.
Frameworks & tools
This skill uses python3 with standard libraries like zipfile and re. It utilizes shell commands including unzip, mdls (macOS), pdfinfo, and strings.
Why this beats prompting it yourself
Generic LLM prompts cannot see the underlying XML structure of a document where tracked changes and metadata live. This skill uses specific regex patterns to inspect the archive content directly, ensuring no hidden data is missed due to the file being "closed."
Use cases
- Auditing a legal contract to ensure internal tracked changes and comments are removed.
- Checking a school essay for invisible zero-width characters used to detect AI generation.
- Cleaning a corporate presentation of internal author names and total editing time.
- Verifying if a PDF contains metadata that links a document to a specific workstation or user.
Known limitations
Removing comments or tracked changes via XML can corrupt files; users must do this manually in their editor. PDF cleaning is limited to basic metadata strings.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Requires python3 and unzip. Cannot programmatically remove comments or tracked changes without risking file corruption. PDF analysis relies on available system tools like pdfinfo or mdls.
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
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 8 days ago
- Free with an Agensi account
- Software
Python 3unzip