Check a 'blacked out' PDF before you send it: have an AI read the file for hidden personal data, then redact it properly and check the new copy again: AI agent prompt

Covering text in a PDF with black boxes often hides it only on screen. On Nov 1, 2026, an X user posted that he had blacked out his name, home address, bank account number and an amount in a contract PDF, saved and reopened it, and everything looked hidden. When he asked Claude Code to check the file for personal data, it returned all of it: the black rectangles were separate objects sitting on top of text that was still inside the file. He had checked every page and nearly sent it. Agents, landlords and property managers send contracts, leases, offers, bank letters and IDs every day, so this check is worth running on anything you've marked up. Use your PDF editor's real Redact tool, apply it, and check the resulting file again before sharing.

SecurityRedacted PDF leak check before sendingbeginner

The prompt

1. Before sending any PDF you have blacked out or covered (contract, lease, offer, bank letter, pay stub, ID), keep the original and work on a copy.
2. Give the copy to an AI that can read files (Claude, ChatGPT or Claude Code) and ask: "Extract all the text in this PDF, including any text hidden under shapes, images or annotations, and any metadata. List every piece of personal or financial information you find: names, addresses, phone numbers, emails, account and routing numbers, ID or tax numbers, signatures, amounts. Tell me which page each one is on."
3. If it finds anything you meant to hide, open the original in a PDF editor with a true Redact tool (not a drawing or highlight tool), mark the items, and apply or save the redaction so the text is removed, not covered.
4. Run step 2 again on the newly redacted file. Send it only when the AI finds nothing you meant to hide.
5. Ask the AI for a one-line checklist you can keep next to your desk: "Redact, apply, re-check, then send."
6. Don't upload documents with sensitive data to an AI tool your company hasn't approved; if you're unsure, ask whoever handles your office's data rules.
Open in ChatGPT Open in Claude Perplexity Download as agent skill (SKILL.md)

How to use it

  1. Copy the prompt above.
  2. Replace every [bracket] with your own details (or fill in the blanks on AgentPlaybook Real Estate, which remembers your answers for the next prompt).
  3. Paste it into Claude, Claude Code, ChatGPT and review the answer before you use it.

Works with: Claude, Claude Code, ChatGPT, PDF editor with a redact tool

Real-world example: x.com/shiqway92/status/2116952031692751228

Featured in: Free AI agent prompts for small business (beginner-friendly)

Before you use it: what to check

Already built into this prompt

Check yourself

Using an AI agent? Download this prompt as a SKILL.md agent skill with these checks built in, or get the .zip for Claude.ai. How to install skills.

Share:Post on XLinkedInFacebookRedditWhatsApp
</> Embed this prompt on your site

Paste this HTML into your blog or website. It's a plain card with no tracking scripts that links back to this prompt.

Want to make this prompt even better? Grade it with the free prompt grader or build your own with the AI prompt generator.

Get the free Solo Founder Stack pack

32 copy-paste agent prompts (PDF) plus 5 new prompts every week.

Get it free

More Security prompts

Draft a malware reverse-engineering briefMalware analysis · intermediateDecide which accounts your AI agent may log into and which actions still need your OK, before you hand over any passwordsAgent login permissions · beginner · worked exampleRun a 15-minute personal security checkup: let an AI interview you about your email, bank, phone and connected AI agents, then hand you a ranked fix listPersonal account security audit · beginnerTake the server your AI agent runs on off the public internet: put it behind Tailscale or a Cloudflare Tunnel so only you can reach itAgent server lockdown · intermediateKeep the computer your AI agents run on patched every morning: one bot watches for new security advisories, another checks what's installed, applies only verified small updates and holds anything bigger for your OKPatch management · intermediateHunt for security bugs like a team: give Claude Code a playbook built from 700+ real disclosed reports, and only test targets you're authorized to testBug bounty & testing · advanced

See all Security prompts →

Related prompts from other teams

Support ticket triageCustomer Support · beginner · worked exampleDraft replies from your FAQCustomer Support · beginner · worked exampleLead qualifierSales · beginnerAccount research briefSales · beginner