claude skills / Sales & Prospecting Skills / audit-first-email / en français

Audit-first email.

The first email is not about you: it gives the prospect an audit of their own website.

See what it produces
  • Claude.ai
  • Claude Code
  • Gmail
  • Outlook
  • PageSpeed Insights
  • Google Rich Results Test
2scores out of 100: AI readability and search basics
3issues, each one reproduced twice before sending
90 to 150words in the email body, a single question at the end

orders of magnitude, depending on your volume

the problem

What gets in the way today.

  • 01Your cold emails look like everyone else's and get ignored from the subject line.
  • 02You promise "more visibility" with nothing the prospect can check in two minutes.
  • 03Some emails claim "ChatGPT doesn't cite you", a sentence nobody can prove and one that discredits the sender.

input, output

You give this. You get this.

What you give Claude: the account brief, your signature and the batch size

Account brief: Harbor & Pine Dental (from account-research, priority A)
Decision maker: Dr. Amara Quinlan, owner, confirmed 4 Nov 2026
Why now: second practice opening in Westmoor, January 2027
Sender: Jordan Ellery, Head of Growth, Northlight Web
Offer: websites and search for dental and medical practices in the region
Proof allowed: none
Recipient language: English | Batch: 5 emails

$ claude "Audit harborpinedental.example and write the first email for Dr. Quinlan."

Audit-first email, ready for approvalexample

AUDIT (extract)

AI readability 58/100 | Search basics 61/100 | audited 4 Nov 2026

A2 Text in the HTML 6/12: treatments list loaded by a script, missing from the source

A4 Organization data 0/10: no JSON-LD found

A5 One page per offer 5/10: no page for implants or orthodontics

B9 Conversion path 0/12: booking errors on mobile, reproduced on two phones, two networks

Subject: AI readability of harborpinedental.example: 58/100

Hello Dr. Quinlan,

We ran the checks that AI assistants and search engines apply when they read a site on harborpinedental.example, on 4 November.

AI readability: 58/100

Search basics: 61/100

This score measures whether AI assistants can read and understand your site. Their answers, on the other hand, vary from one assistant to another.

What is blocking:

- Booking fails on a phone: the "Book an appointment" button leads to an error page on mobile.

- No page for implants: your implant ads land on a homepage that does not mention them.

- Westmoor is only in an image: assistants and search engines cannot read the new location.

We are Northlight Web, a web and search agency for dental and medical practices in the region. With Westmoor opening in January, we would fix booking first.

Would you like the full audit, with the fix for each point?

Jordan Ellery

Head of Growth, Northlight Web

If this is not relevant, reply "no" and we will not write again.

Body: 147 words | Planned: Tuesday 11 November, 9:20, recipient's time zone | Status: awaiting approval

the method

What Claude does, in order.

  1. step 1Run the audit

    Both 100-point grids run on the homepage and the page that makes money. Each check is recorded with its evidence and date, and the raw sheet is kept.

  2. step 2Pick 3 issues

    Among the failures, Claude keeps the three that weigh most on revenue, then checks them again on another device. An issue that cannot be reproduced is dropped.

  3. step 3Write the subject and body

    The score becomes the subject, the three issues the body, one line says who you are, and a no-commitment question ends the email.

  4. step 4Review and schedule

    Words counted, checklist, then one review page for the whole batch. Sends Monday to Wednesday, 9 to 10 am in the recipient's time zone, 5 per day.

demo in French / click, try it

Your turn.

An interactive demo of the skill. It is in French, the pack works in any language: ask Claude to answer in English.

objet : AI readability of harborpinedental.example: 58/100

the film / 44 seconds, with sound

The pack film.

the pack film

In 8 seconds.

the guardrails

What the skill never does.

  • We write "AI readability", never "AI visibility": the score measures whether AI can read the site, not whether it cites it.
  • Every score comes from the grid, applied to the live site on a given date: a prospect who reruns the tests gets the same result.
  • No invented proof: a client name or a result only appears if you provided it and are allowed to use it.
  • Claude drafts and schedules, it sends nothing: a person approves every email and every batch.

Questions

Why put the score in the subject line?

Because a number about your own website gets read before the sender's name. No question mark, no emoji, no first name in the subject.

What if a cookie banner or a maintenance page skews the audit?

Claude scores the real page, or does not send. An issue it cannot reproduce a second time is removed from the email.

Does the skill send the emails?

No. It delivers a review page with each email exactly as it will go out and its planned time. You approve, then schedule in your email client.

Is it compliant with cold outreach law?

The skill requires the sender's identity, an opt-out sentence and role-based business addresses. Rules vary by the recipient's country, so it asks you to check when in doubt.

the pack

In the pack: Sales & Prospecting Skills.