This AI prompt gives you honest feedback on your work before you send it. It names the 3 weakest parts, worst first, with a solution for each. Then it tells you the one part that’s working, a score out of 10 & the single change that would raise it most.

It works on an email, a resume, a pitch or a landing page.

The prompt for honest feedback before you send anything

Here’s the one I reach for most, a blunt critique of anything before I send it.

Act as a demanding {{role, e.g. senior editor / hiring manager}} reviewing my {{thing}}.

What it's for: {{goal}}.
Who it's for: {{audience}}.

Give me:
- The 3 weakest things, worst first, with a specific fix for each.
- One thing that's working, so I don't break it.
- A score out of 10 and the single change that would raise it most.

Be blunt. I want it better, not validated.

WORK: {{paste it}}

Why it works: Asking for the 3 weakest things, worst first, makes the model rank its notes instead of listing everything it could say. “One thing that’s working” stops it from rewriting the part you got right. And “I want it better, not validated” tells it to be honest instead of polite.

One gap: The prompt is only as good as what you put in the blanks. Fill {{thing}} with “my email” & the feedback is so generic it fits any email.

Simple fix: Get specific. Fill it with “the opening 2 lines of my cold email to a VP” & the notes get sharp. “My email” gets a shrug.

Want to go deeper: Once you know what’s weak, this prompt rewrites it in your voice & gives you 2 versions, one safe and one bolder.

Rewrite this {{content type, e.g. email / bio / paragraph}} to be more {{quality, e.g. concise / warm / direct}}.
Keep my meaning and my voice. Don't invent facts.
Give me 2 versions, one safe and one bolder, with a one-line note on what changed.

TEXT: {{paste it}}

Original post by u/Ok_Negotiation_2587. Link

Where to take the feedback next

To see how a piece you admire was built, the writing style prompt breaks it down line by line.

This prompt first ran in issue 5, next to an angry message prompt for the replies you don’t want to get wrong.