I want to start today with a small heresy. Most prompt libraries are trash.
Not because the prompts themselves are bad. Because they are presented as if a great prompt is a magic spell you copy once and use forever. That is not how this works. A great prompt is a tool that fits a specific job in your workflow. The prompt that revolutionizes my Tuesday morning will probably do nothing for yours, because your Tuesday morning is shaped differently.
So instead of giving you a hundred clever prompts you will save and never use, I am going to do something more useful. I am going to give you seven prompts I personally run every week, the exact wording I use, the place each one fits in my actual workflow, and the principle behind each one so you can adapt it to your own.
Steal what fits. Modify what almost fits. Skip what does not apply to you. Then build your own vault from the same principles.
A quick note before we get into it. I run most of these in Claude because the longer reasoning works well for the kind of analytical and writing work I do. A few of them I switch to ChatGPT when I want a faster, punchier output. Galaxy.ai lets me bounce between models without juggling subscriptions, which is useful if you want to test the same prompt across multiple engines and pick the winner.
Onward.
Prompt One: The Monday Triage
When I run it: First thing Monday morning, before I touch email.
What it does: Takes my calendar, my outstanding to-do list, and any open client threads from the previous week, and produces a ranked priority view for the week ahead.
The wording: "Below is my calendar for the week, my current open to-do list, and a brief summary of my client thread status. Help me build a ranked priority view for the week. Identify the three highest-leverage moves I can make, the two things that look urgent but are not, and the one thing I keep deferring that I should either commit to or formally drop. Be direct. Do not hedge."
Why it works: The prompt forces a separation between urgent and important. It also names the deferred item explicitly, which is the thing my brain is trying to hide from me. The instruction to "be direct" matters more than people realize. Without it, you get hedging. With it, you get a useful answer.
Adapt it by: Replacing "client thread status" with whatever your operational signals are. Sales pipeline. Project deadlines. Team workload. The structure stays. The inputs change.
Prompt Two: The Cold Reader
When I run it: Whenever I have written something I am about to send to a real audience and I want a second pair of eyes that is not my own.
What it does: Reads my draft as a skeptical version of my target reader and gives me back the three weakest moments in the piece.
The wording: "You are reading the following piece as a busy small business owner who is skimming on a phone between meetings. You are friendly to the topic but skeptical of generic advice. Identify the three weakest moments in this piece. For each one, tell me specifically what the problem is and what would make it stronger. Do not soften your feedback to be polite."
Why it works: The persona forces a specific lens. The instruction to be unpolite removes the AI default of validating everything you do. The "three weakest moments" framing constrains the output to be useful instead of overwhelming.
Adapt it by: Changing the persona to match your actual audience. A CFO reviewing a proposal. A potential client reviewing your sales page. A junior team member trying to follow your instructions.
Prompt Three: The Meeting Distillation
When I run it: After any client meeting or strategic call. I use Fathom to capture the transcript, then drop the transcript into this prompt.
What it does: Pulls the actual decisions, the actual action items, and the actual unresolved questions out of a meeting transcript, separating signal from chitchat.
The wording: "Below is the transcript of a meeting. Produce three sections. First, decisions made, with who is accountable for each. Second, action items committed to, with the owner and the deadline if one was named. Third, unresolved questions or topics deferred, with a short note on why each was set aside. Ignore pleasantries, throat-clearing, and tangents that did not lead anywhere. Be specific."
Why it works: The three-section structure mirrors how a useful meeting recap should be organized. The "ignore pleasantries" instruction is what separates this from the default summary, which tends to give equal weight to "we caught up about the weekend" and "we agreed to ship by Friday."
Adapt it by: If your meetings tend to surface customer insights, add a fourth section called "customer signals worth flagging." If they involve sales calls, add a section for "objections raised and how they were handled."
Prompt Four: The Offer Stress Test
When I run it: Whenever I am about to launch a new offer, sales page, or pricing change.
What it does: Pressure-tests the offer from three angles before it goes live.
The wording: "Below is the description of an offer I am considering launching. Pressure test it from three angles. First, the buyer who is exactly right for this and ready to say yes. What do they love about it, and what would push them to act today? Second, the buyer who almost fits but has one specific objection. What is the objection, and how should I answer it directly in the offer page? Third, the buyer who is not the right fit at all. How can I make that obvious so I do not waste their time or mine? Be honest. Do not be a cheerleader."
Why it works: Three buyer personas in one pass. The "do not be a cheerleader" instruction stops the validation reflex. The third persona, the wrong-fit buyer, is the one most owners forget about and the one that saves the most refund headaches.
Adapt it by: Adding a fourth angle if you have specific competitors. "Fourth, a buyer who is comparing this to [competitor]. What are they thinking, and how should I position against the comparison without trash-talking?"
Prompt Five: The Email I Do Not Want To Write
When I run it: When I am procrastinating on a hard email. The follow-up to a quiet client. The push back on a partner. The decline of an opportunity. The honest update when the news is not good.
What it does: Writes the first draft of the email in a tone I can edit, removing the blank-page paralysis.
The wording: "I need to send the following email. The situation is [brief context]. The outcome I want is [specific outcome]. The tone should be direct, professional, warm where appropriate, and free of any unnecessary apology or hedging. Do not use the word 'just.' Do not start with 'I hope this email finds you well.' Do not include filler. Draft the email."
Why it works: Specifying the outcome focuses the draft on the actual goal. The list of forbidden moves keeps it out of the generic professional-email zone. The "do not include filler" instruction is doing more work than people think.
Adapt it by: Adding a constraint on length. "Keep it under 100 words." Or specifying a relationship history. "We have worked together for three years and have a casual rapport."
Prompt Six: The Weekly Win Audit
When I run it: Friday afternoon, before I close the laptop for the weekend.
What it does: Forces a quick written reflection on what actually moved this week, not just what felt busy.
The wording: "Below is a summary of what I worked on this week, including major calls, deliverables shipped, and decisions made. Help me identify three things. First, the highest-leverage thing that happened this week, even if it felt small. Second, the thing I spent the most time on that produced the least result. Third, the lesson I should carry into next week. Be honest. The point is to learn, not to feel good."
Why it works: The "highest leverage" framing separates impact from volume. Most weeks the most important thing was a five-minute conversation, not the four-hour project. The "least result" question is the one nobody wants to ask themselves, which is exactly why it pays.
Adapt it by: If you run a team, do this prompt with team-level inputs and use the output as the basis for a Friday team sync. Done weekly, it will reshape how the team thinks about effort versus impact.
Prompt Seven: The Blank Page Breaker
When I run it: Anytime I sit down to start a new piece of writing, a new offer outline, a new strategy doc, anything where the first paragraph is the hardest part.
What it does: Generates five different opening angles so I can pick one and run.
The wording: "I am about to write a piece on [topic]. Generate five different opening angles for the piece. Each one should take a distinctly different approach. One should be a contrarian take. One should be a personal story. One should be a sharp question. One should be a specific example or case. One should be a confident statement of a counterintuitive truth. For each, give me just the opening paragraph, no more. I want to feel pulled into the piece by the time I finish reading it."
Why it works: Five angles means I am choosing, not generating. Choosing is faster than generating from scratch. The required variety in approach forces the model to stop defaulting to the same opening shape every time.
Adapt it by: Asking for ten openings if you want more options. Specifying the publication context. "These are for a daily newsletter that targets entrepreneurs."
The principles behind the vault
If you only take the prompts and leave the principles, you got a fraction of the value. Here is the operating system underneath.
Specificity beats cleverness. The prompts that work are not clever. They are precise. They name the persona, the constraint, the format, the forbidden moves. Vague prompts get vague answers. The best prompt engineering skill is the willingness to spell out exactly what you want.
Forbidden moves matter. Telling the model what to avoid is often more powerful than telling it what to do. "Do not hedge" produces a different answer than "be confident." "Do not include filler" produces tighter writing than "be concise." Use the negative space.
Personas focus the lens. Asking for analysis from a specific reader's perspective produces sharper feedback than asking for analysis in general. The right persona is usually a specific person you can almost picture.
The output format is part of the prompt. If you want three sections, ask for three sections. If you want a ranked list, ask for a ranked list. The model will fill whatever shape you give it. Choose the shape on purpose.
Iteration is fine. The first run rarely nails it. Adjust the prompt based on what was off. Run again. The work is in the dialogue, not in the perfect first draft.
Where this goes next
If you want a deeper library of prompts organized by use case, the AI Workflow Blueprint includes a prompt vault with over fifty production-ready prompts for common small business workflows, all written with the principles above and tested in real operations. Reply with BLUEPRINT and I will send it over. Forty seven dollars, and the prompts alone will pay for it within a week if you actually use them.
If you want the full system for building a custom prompt library tied to your specific business processes, reply with ACCELERATOR. The AI Business Accelerator at ninety seven dollars walks through the structure, the documentation, and the team handoff that makes a prompt library useful across more than one person.
Steal a prompt this week. Run it three times. Make it yours.
Jordan Hale
The AI Newsroom