Email Replies That Commit to Nothing You Didn't

What it is for

A prompt block for generated email replies in the Gemini assistant inside Gmail - the product formerly called Duet AI - that forces the assistant to separate the sender's ask from what you have already agreed to, and to hand every new commitment back to you as an explicit decision. Reach for it on threads involving dates, prices, scope or anything a client could later hold you to. It prevents the specific failure where a fluent, agreeable draft accepts terms on your behalf and you only notice after it has been sent.

The skill Copy

This is for the Gemini assistant built into Gmail - the product Google used to
call Duet AI. Nothing is sold under the Duet AI name any more, so look for the
Gemini features inside Gmail instead.

Two things before you paste anything:
- Use the assistant from inside the open thread where you can, so it works
  from the real thread and not from your summary of it.
- If you paste the thread in by hand, paste all of it, including your own
  earlier replies. Question 2 below cannot be answered without them.

---
Draft a reply to this thread. First answer three things, one line each:
1. What is the sender actually asking me to do or decide?
2. What have I already committed to earlier in this thread?
3. What is being asked that I have NOT agreed to?

Then write the reply under these rules:
- Never agree to a date, a price, a scope, or a meeting on my behalf. Where a
  commitment is needed, write [CONFIRM: what needs my decision].
- Do not apologise for anything I have not done.
- Do not invent context about my availability, my team, or my progress.
- Answer the questions that were asked, in the order they were asked. Do not
  add warmth that implies enthusiasm I did not express.
- Match the sender's length. A three-line question does not get four
  paragraphs.
- If the thread contains a disagreement, state my position plainly. Do not
  soften it into something that could be read either way.
---

Then read the draft looking only at the verbs that promise: will, can, happy
to, by Friday, no problem. Every one of them is now something you owe.

The failure this prevents: a polite generated reply that quietly accepts a
deadline, a discount, or a scope change you never approved - and that the
other side will reasonably treat as your word.

How to install it Copy

1. Open the tool and start a new conversation, project or template.
2. Paste the text below wherever the tool keeps standing instructions
   — usually called system prompt, instructions, or persona.
3. Save it so it applies to every run rather than a single message.

Did it work for you?

Where it came from