Mask before Send
Before your agent sends text to an outside service, it swaps e-mails, phones, IBANs, card numbers, IDs and any values you name for placeholders, keeps the map on your machine, and restores the real values in the reply.
$5
Works with the AI Agents you already use
Mask before Send
Example session with this skill installed
Before you send note.txt to the external summariser, follow mask-before-send. The person named in it is Jane Roe.
- Read your context and instructions
- Compiled the mask before send
Example session. The script output below is real (file paths shortened); the note and the reply text are made up for the example.
1. See what the rules find
$ node scripts/redact.mjs check note.txt
3 identifier(s) found by rules:
- EMAIL: jane.roe@example.org
- PHONE: +44 7700 900123
- IBAN: GB33BUKB20201555555555
2. Mask, adding the name the rules cannot see
$ node scripts/redact.mjs mask note.txt --also "Jane Roe"
4 value(s) masked.
Mail [EMAIL_1] or call [PHONE_1]. IBAN [IBAN_1]. Signed, [CUSTOM_1].
| Placeholder | Type | Stays on your machine |
|---|---|---|
| [EMAIL_1] | jane.roe@example.org | |
| [PHONE_1] | phone | +44 7700 900123 |
| [IBAN_1] | IBAN | GB33BUKB20201555555555 |
| [CUSTOM_1] | named by you | Jane Roe |
Only the masked line is sent to the external service.
3. Restore the values in the reply
$ node scripts/redact.mjs unmask reply.txt --name note
Warning: placeholders not in the map (left as they are): [ORG_3]
Thanks Jane Roe, we will e-mail jane.roe@example.org and call +44 7700 900123. Ref [ORG_3].
Report to the user: 4 values were masked before sending (1 e-mail, 1 phone, 1 IBAN, 1 name you gave me). The reply mentioned a placeholder that was never in the document, [ORG_3], so I left it as it is. The map is in .redact/note.map.json; keep .redact/ out of git.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Agents paste logs, tickets and letters into web searches, third-party APIs and other AI services. The task rarely needs the real e-mail address, phone number or bank account, yet those go out with the text.
What this skill does
It gives your agent a rule and a small local tool:
redact.mjs check <file>lists the identifiers the rules find.redact.mjs mask <file> --also "<name or address>"replaces them with placeholders such as[EMAIL_1]and saves the map in.redact/on your machine.- The agent sends only the masked text.
redact.mjs unmask <reply> --name <id>puts the real values back into the answer and warns about placeholders the reply invented.
What the rules detect
E-mail addresses, IBANs (checksum-validated), payment card numbers (Luhn-validated), Turkish national IDs (check digits), US SSN format, UK National Insurance numbers, phone numbers and UK postcodes. Validation means a random 16-digit number is not flagged as a card.
What is inside
SKILL.md: when to use it, the steps and the rules.scripts/redact.mjsand two small modules inscripts/lib/. Node only, no dependencies, no network calls.
Limits
- Rules do not detect names, street addresses or organisations. The agent has to read the text and pass those with
--also; anything it misses is sent. - ID formats are limited to the ones listed above.
- It lowers the risk of leaking personal data. It is not a compliance guarantee.
- Needs Node 18 or newer. New skill: no usage numbers yet.
The same masking code runs in our open-source demo app: github.com/ahmetilhanli/redact-and-ask
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
Rules do not detect names, addresses or organisations: pass them with --also. ID formats limited to Turkish national ID, US SSN format and UK NI number. Lowers risk; not a compliance guarantee. Needs Node 18+.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 2 days ago