Free tools

AI prompt generator for sales reps

Call prep, call summary, follow-up email, objection handling: pick the use case, fill in the blanks, and copy a solid prompt into ChatGPT or Claude.

No made-up facts allowed

1Pick a use case

Prepare

Write to the prospect

After the call

2Fill in the blanks

Prep a cold call

Example filled in: replace the highlighted values with yours.

required
required

The name plus whatever helps: industry, size, location.

required

One concrete sentence: for whom, what, which problem. Typed once, reused in every use case.

Optional. The more specific the context, the less generic the answer.

Optional. Separate them with semicolons.

Answer settings

Tone

3Your prompt

Prompt ready3/3
## Role You are a B2B cold calling coach. You prepare reps for short cold calls that are useful to the prospect and free of pressure. ## Context I sell field service scheduling software for HVAC and maintenance contractors: technician dispatch, mobile work orders, and service contract tracking. I'm about to call Dana Reyes, VP of Operations at Northwind Mechanical (commercial HVAC contractor, about 120 technicians, Denver area). Call goal: book a 20 to 30 minute meeting. What I already know: <context> Northwind is hiring 8 service technicians, per its careers page. Dispatch still seems to run on spreadsheets and phone calls, based on a former employee's review. </context> Objections I hear a lot: we already have a tool; bad timing; just send me an email ## Task Build my call sheet. It has to fit on one screen and be readable at a glance while the phone rings. ## Output format 1. **Opener** (two sentences max): who I am and why I'm calling this person, today. 2. **Reason for the call**: one likely challenge for this role, tied to my offer, phrased as a question rather than a claim. 3. **Five discovery questions**, open-ended, from easiest to most committing. 4. **Three likely objections**, each with a response of 30 words max: acknowledge, ask a question, offer another angle. 5. **Close**: the next step to propose, with two specific time slots. 6. **If I hit voicemail**: a 20-second message. ## Constraints - Written to be spoken: short sentences, plain words, no jargon. - No claim about the company or the person that isn't in my information. - Write your answer in American English. - Direct, plain tone: short sentences, no superlatives, no filler. ## Accuracy rules - Do not make up facts, numbers, customer names, quotes, or news. Use only the information in this message. - If useful information is missing, don't guess: write [To verify] where it belongs, or flag an inference with [Assumption]. - If my request looks off (wrong target, wrong timing, wrong channel), say so in one sentence before answering. - Don't ask me questions: answer directly and flag your assumptions.
Your inputsSettings2,111 characters · 355 words

Click a highlighted value to edit its field.

Open in ChatGPTOpen in Claude

The link keeps the use case, the settings and your offer. Never names, notes, or transcripts.

Then, in the same chat

Click to copy, then paste it after the first answer.

Get more out of the answer

  • Read the sheet out loud once: any sentence you trip over will be worse on the phone. Ask "simplify line 2."
  • To rehearse the whole call before you dial, switch to the "Role-play a sales call" use case: your prospect and offer carry over.

The call summary, without the prompt. Skipcall transcribes and summarizes your prospecting calls with AI and files them with the contact, so the notes are there when you write the follow-up.

Create an accountSee transcription and AI summaries

Everything stays in your browser: nothing you type is sent to Skipcall. How it works

Prep a cold call

Prompt ready

What makes a good sales prompt

Ask ChatGPT to “write a cold email for my software” and you get the same email everyone else gets: vague, flattering, and full of claims about your product it cannot know. The model is not the problem. It simply had nothing to work with.

A good prompt gives the AI what a new hire would need on day one: who it is playing, the context of the deal, the exact task, the shape of the answer, and the rules it must not break. The generator writes those parts for you, from a few fields you fill in.

  1. 1

    ## Role

    Who the AI is playing: a cold calling coach, a research analyst, a sales leader.

  2. 2

    ## Context

    Your offer, your prospect, what you already know. Long inputs go first, inside tags.

  3. 3

    ## Task

    One clear job, phrased as an instruction.

  4. 4

    ## Output format

    Numbered sections, word limits, tables: an answer you can use without editing.

  5. 5

    ## Constraints

    Written to be spoken, no filler, language, tone, and register.

  6. 6

    ## Accuracy rules

    No made-up facts, gaps marked [To verify], assumptions flagged as such.

What these prompts are built on

The structure follows the prompting guides published by the two companies behind ChatGPT and Claude (checked November 6, 2026). They agree on the essentials.

  • Prompting best practices Anthropic

    Be clear and direct, explain why an instruction matters, show an example of what you want, separate instructions from inputs with XML tags, put long documents at the top, and ask the model to ground its answer in quotes.

  • Prompt engineering guide OpenAI

    Organize the prompt into identity, instructions, examples, and context, with Markdown headings and XML tags to mark where each input starts and ends.

The generator adds one thing those guides leave to you: rules written for sales, like “no promise that is not in the notes” in a follow-up email, or “no proof provided, invent none” in a cold email.

The 13 prompts and when to use them

Each link opens the generator on that use case.

39 follow-up prompts to chain after the first answer

The first answer is a draft. These one-liners are what sales reps type next, in the same conversation, so the AI keeps all the context. Click to copy.

Prep a cold call
Role-play a sales call
Research an account
Prep a discovery call
Build an account plan
Cold email
LinkedIn message
Voicemail
Follow up with a silent prospect
Call summary
Follow-up email
Handle an objection
Analyze a call transcript

How to get an answer you can actually use

A prompt is the start of a conversation, not a vending machine. Five steps make the difference.

  1. 1

    Fill in what you really know

    Three precise lines beat ten vague ones. An empty optional field is better than a guess: the prompt then tells the AI what it does not know.

  2. 2

    Paste, then read before you use it

    Look for [To verify] and [Assumption] first. They mark exactly where the AI is missing information, which is where you should check.

  3. 3

    Iterate in the same conversation

    “Shorter,” “more direct,” “version 2 but with the hook from version 3”: the AI keeps the context, so refining costs one line. Each use case suggests three follow-ups under the prompt.

  4. 4

    Show it what good looks like

    In the writing use cases, paste a message of yours that got a reply. The AI matches its tone and length instead of producing the generic voice everyone recognizes.

  5. 5

    Keep your best prompts

    Download the .txt or copy the share link with your offer already filled in. Your team starts from the same base, not from a blank page.

What you can paste into ChatGPT or Claude

Call notes and transcripts contain personal data: names, phone numbers, sometimes things people said in confidence. Pasting them into a consumer AI assistant means sending them to that provider.

France’s data protection authority (CNIL) published a Q&A on using generative AI on July 18, 2024. Its recommendations apply well beyond France: users should only submit information they are allowed to share, organizations should limit the personal data entered in prompts and may forbid some categories, outputs should be checked rather than used as is, and suggested prompt templates help frame usage.

Whether your conversations train the provider’s models depends on your plan. At OpenAI, conversations from personal ChatGPT plans can be used for training unless you turn off “Improve the model for everyone” in Data Controls, while Business and Enterprise workspaces are not used for training by default. At Anthropic, Claude Free, Pro, and Max users choose in their privacy settings whether their chats are used for training; Claude for Work and the API are not covered by that setting.

Fine to paste

  • Your offer, your pitch, your objections
  • Job titles rather than names
  • Notes with names replaced by “the prospect”
  • Public company information

Keep out

  • Sensitive data: health, political, religious or union views
  • Email addresses, phone numbers, bank details
  • Anything a customer told you in confidence
  • Confidential deal terms your company forbids sharing

Information, not legal advice. Provider settings change often: check yours before pasting customer data (checked November 6, 2026).

Common mistakes

  • A one-line prompt

    “Write a follow-up email” forces the AI to invent the context. It will, convincingly, and wrongly.

  • Letting the AI invent proof

    A customer name or a percentage the AI made up, sent to a prospect, is worse than no proof at all. Leave the field empty: the prompt then forbids it.

  • Sending without reading

    Dates, names, and promises are where AI slips most. Read every email before it goes out, out loud if it is meant to be spoken.

  • Pasting the whole CRM record

    More data is not more context. Paste what matters for this task, and nothing personal you do not need.

  • Asking for one version

    Three versions with three angles let you compare. The generator asks for them for cold emails, LinkedIn, and follow-ups (two versions for voicemails).

Keep going

The tools that pair with this generator, and our guides on AI in sales.

Nothing you type leaves your browser

The prompt is assembled in your browser from templates written by hand. No AI runs on this page, nothing you type is sent to Skipcall or anyone else, and nothing is stored. The AI only receives the prompt when you paste it into your own assistant.

The “Open in ChatGPT” and “Open in Claude” buttons copy the prompt, then open the assistant with the prompt in the address so it can be prefilled. That address parameter is not officially documented by OpenAI or Anthropic: if the box comes up empty, paste. The share link carries only the use case, the settings, and the fields that describe your offer.

A calendar full of meetings starts with a call.

Get started now Book a demo

No commitment · Number live in 2 minutes

Frequently asked questions

Yes. They are plain text with Markdown headings and a few XML tags, a format all major assistants read well. The “Open in” buttons cover ChatGPT and Claude; for any other assistant, copy and paste.
Free, no signup, no account. You run the prompt in your own assistant, on your own plan.
Because each section removes a guess the AI would otherwise make: who it is playing, what you sell, what format you want, what it must not invent. A long prompt costs nothing to copy, and saves you three rounds of corrections.
Yes, that is what the “Analyze a call transcript” use case is for. Replace names with “Rep” and “Prospect” first, and use the mask button to remove emails, phone numbers, and IBANs. Never paste sensitive data.
It makes up far less, not nothing. The prompts forbid invented facts and ask the AI to mark gaps with [To verify] and inferences with [Assumption], which tells you where to look. Always reread before sending.
Prefilling relies on an address parameter neither OpenAI nor Anthropic officially documents, and very long prompts are not prefilled. The prompt is copied before the tab opens: paste it with Ctrl + V or Cmd + V.
Yes. The share link opens the same use case with the same settings and your offer already filled in. Names, notes, and transcripts are never in the link. You can also download the prompt as a .txt file.
Use the “Role-play a sales call” use case. The AI plays the prospect one line at a time, at the difficulty you pick, and raises the objections you list. Type END to get a score, the two key moments with your exact lines, and one thing to work on. Type PAUSE to ask a question out of character.
Here, your offer carries over from one use case to the next, and the share link keeps it. In your assistant, you can also save a short description of your offer in its custom instructions or in a project, if your plan offers them, so every conversation starts with it.