Email Rewrite — From Vague to Shippable

Hagar's first real test. Her manager asked her to draft an email to her landlord about extending her lease. She tried "Write me an email" and got a fillable template. We saw that in Module 1. Now we use the skeleton to write the email properly — and watch the same model produce something Hagar can copy-paste in one go.

The before

Reminder of what "Write me an email." produced:

# Email Template

**To:** [Recipient's email]

**From:** [Your email]

**Subject:** [Your subject line]

---

Dear [Recipient's name],

I hope this email finds you well.

[Main body of your message goes here...]

It's a template, not an email. Useless for a same-day landlord conversation.

The after — the skeleton applied

Now Hagar writes the same prompt with the five-slot anatomy in mind.

Write a 90-word email to my landlord asking to extend my lease by 6 months.
Tone: warm and professional. I have lived here 2 years, paid rent on time,
and have a stable job. Sign as "Sara". Output only the email body, no subject line.

Captured output:

Dear [Landlord's Name],

I hope this message finds you well. I've thoroughly enjoyed living here for the past two years and would love to extend my lease for an additional six months.

As you know, I've consistently paid rent on time and maintained the property with care. My job remains stable, ensuring continued reliable tenancy.

This extension would be ideal for my current situation, and I'm happy to discuss any terms or paperwork needed.

Thank you for considering my request. I appreciate your understanding and look forward to hearing from you.

Warm regards,
Sara

Captured from Claude Sonnet 4.5 (claude-sonnet-4-5) on 2026-04-27. Re-runs may differ slightly.

This time the model produced a real, signed, paste-ready email. Same model, same task, completely different result.

Mapping the prompt to the slots

Hagar's prompt didn't use the labelled Role: / Task: form, but the skeleton is still all there:

Slot	Where it appears
Role	Implicit — a polite tenant. Could have been more explicit but the tone instruction did the heavy lifting.
Task	"Write a 90-word email to my landlord asking to extend my lease by 6 months."
Context	"I have lived here 2 years, paid rent on time, and have a stable job."
Input	(none — the email is generated from the context, not from a source paragraph)
Output	"Tone: warm and professional. ... Sign as 'Sara'. Output only the email body, no subject line."

The output spec did three things at once: locked the tone, locked the signature, locked the format ("body only, no subject line"). Without the last instruction, the model would have produced a Subject: line that Hagar would have had to delete.

What still needs human polish

Look closely. The output has [Landlord's Name] as a placeholder. The model didn't have the actual name. If Hagar's flow is "write email → fill name → send", she's done in 30 seconds.

This is the realistic shape of "ship-ready" — not "zero edits", but "zero rewrites". You replace one placeholder. You don't restructure paragraphs.

A pattern Hagar can save

The shape Hagar just used is reusable. She can save it in a notes app:

Write a [LENGTH] email to [RECIPIENT] asking [GOAL].
Tone: [TONE]. [3 BULLETS OF CONTEXT.]
Sign as "[NAME]". Output only the email body, no subject line.

Tomorrow she'll fill those in for a different email — a follow-up to a recruiter, a thank-you to a vendor, a polite chase to a colleague. Same shape, different content.

This is how prompt engineering actually compounds in daily work: not by inventing new prompts every time, but by saving good shapes and adapting them.

Next: meeting summary — same skeleton, different content type. :::