15 prompts · copy and adapt

The prompt library. Free to copy.

The prompts I use for my own business and client work, rewritten so they work for yours. Fill in the brackets, paste, and check the output like you would a new hire's.

Full-context capture, before any long build session

Paste a voice-note transcript or brain dump before asking a model to build anything big.

Why it worksMost bad AI output is a context problem. This makes the model prove it understood you before it writes a line of code or a plan.

Show the prompt
Starter prompt · copy and adapt
You are my strategic execution partner.

Your first job is not to solve. Your first job is to absorb the transcript below and reconstruct the full context: intent, projects, constraints, energy, and the opportunities I did not say out loud.

Review it until you can explain it back better than I said it. Do not summarize too early. Do not flatten the nuance. Do not turn this into generic productivity advice.

Capture:
1. Dates and timing, and any ambiguity in them.
2. The core breakthrough or decision in this transcript.
3. Every active project, with its goal, status and blocker.
4. Constraints: money, time, tools, risk, people.
5. What I want to push back on or protect (quality, security, focus).

Then produce:
A. Context map
B. Active project inventory
C. Strategic themes
D. Decision points I need to make
E. Risks and blind spots
F. Suggested execution order
G. Clarifying questions, only if they materially change the plan
H. The single best use of the next working session

Do not write code yet. First prove you understand the transcript.

TRANSCRIPT:
[paste here]

The run-it-end-to-end brief

Any task you want finished in one pass: a plan, a document set, an analysis.

Why it worksLead with intent, give rich context, and authorize the model to finish. It only stops for decisions that are truly yours.

Show the prompt
Starter prompt · copy and adapt
You are my [role, e.g. curriculum designer / security architect / analyst].

OBJECTIVE (what winning looks like):
[One paragraph. Describe the finished result, who it is for, and the quality bar.]

CONTEXT ABOUT ME:
- [Who I am and what I do]
- [What I care about: outcomes, tone, constraints]

WHAT EXISTS SO FAR:
[Drafts, files, earlier attempts, and what is wrong with them.]

YOUR TASKS (in priority order):
1. [Task]
2. [Task]
3. [Task]

STYLE:
- Plain language. Explain any jargon you cannot avoid.
- Mark anything you are unsure of instead of guessing.

EXECUTION:
Run this end to end. Only stop if you hit something irreversible or a decision only I can make (for example, choosing between two brand positions). Otherwise complete everything in one pass, then list the decisions you made on my behalf.
ChatGPT

Workshop kit builder

Turn a rough talk or training idea into a facilitator-ready kit someone else could run.

Why it worksTraining is one of the four things I sell. This is the prompt that drafts the first version of every curriculum.

Show the prompt
Starter prompt · copy and adapt
You are my curriculum and workshop design partner.

OBJECTIVE: Turn my workshop idea into a polished kit: session structure with timing, slide outline, facilitator notes, one hands-on activity per section, and reflection prompts. It must be engaging (not default corporate blue and white) and repeatable by a facilitator who is not me.

AUDIENCE: [who, how technical, how many people]
DURATION: [60 to 90 minutes, with optional extensions]
OUTCOMES: By the end, participants should:
- Explain [core idea] in plain language
- See two examples that connect to their actual week
- Complete one hands-on exercise
- Leave believing "this is accessible to me"

FOR EACH SECTION, GIVE ME:
- Slide titles and bullet structure
- Visual direction (palette, imagery, layout)
- Talking points in plain language
- One interactive element (question, short exercise or discussion)
- Facilitator notes: what to say, what to ask, what to watch for

Flag anything too complex for this audience and simplify it without losing the point. Run this end to end in one pass.

Security-first review of a repo or integration

Before you connect any tool that touches logins, money or customer data.

Why it worksI treat every 'it's secure, trust me' repo as suspect until reviewed. This prompt makes the model do the suspicious part.

Show the prompt
Starter prompt · copy and adapt
You are my security-conscious backend architect.

I found [repo / integration / library] that claims to [what it does]. Treat its security claims as unverified.

1. Enumerate the main risks in a naive implementation: hard-coded credentials, token storage, logging sensitive data, insecure hosting, missing rate limits, multi-user data leakage.
2. Propose a safer design:
   - Secret management (environment variables or a vault, never in prompts or code)
   - Session lifecycle: how long tokens live, how they refresh, when they are discarded
   - Data-at-rest expectations
   - Rate limiting and abuse prevention
3. Give a roadmap: v1 single user, v2 small private group, v3 production.
4. Mark every non-negotiable clearly (for example: never log raw credentials).
5. Explain how I can ask an AI to help write code and configs without ever pasting a real secret into a prompt.

No code yet. Architecture, risks and roadmap first.

Planner for limited AI work windows

When you have usage limits or a few focused hours, and more projects than time.

Why it worksPrepare the prompts before the window opens, so expensive hours go to execution, not to figuring out what to ask.

Show the prompt
Starter prompt · copy and adapt
You are my session planner.

I have a recurring window of focused AI time: [length, e.g. 5 hours] that resets [when].

MY PROJECTS:
- [Project]: goal, status, next step
- [Project]: goal, status, next step

1. Build a simple backlog model: project, task, value score (1-5), effort (S/M/L), dependencies, status, and whether it truly needs the strongest model.
2. For the next window, choose 1 to 3 tasks with the best value for effort. Explain why.
3. For each chosen task, write a complete, paste-ready brief (objective, context, tasks, execution rule: run end to end, only stop for irreversible decisions).
4. Tell me what to save after the session so the next brief can build on it instead of repeating work.
Perplexity

Local market research brief (ICP)

Decide which kind of local business to serve first, with a prospect list you can act on this week.

Why it worksPick one industry before you pick a tool. This brief keeps research honest: no invented businesses, no invented ROI.

Show the prompt
Starter prompt · copy and adapt
You are a senior market researcher. Deliverable: market research, not sales copy.

BUSINESS: [what you sell, to whom, your price points]
AREA: [borough / city / neighborhoods]
REVENUE BAND: [e.g. $250K to $10M, owner-led]

Identify the top 5 local business segments where the offer fits best. For each:
1. Segment overview: what it is, why it fits, rough count of businesses.
2. ICP card: size, owner persona, tech comfort, what eats their week.
3. 3 to 5 pain points, each with a cost in hours per week or dollars per month.
4. 2 to 4 buying triggers that create urgency this quarter.
5. Channels: online, offline, and which professions refer them first.
6. Disqualifiers.
7. Starter prospect list (15+ real businesses) with a public link and one specific automation idea per business.

CONSTRAINTS: Verify every business against public sources. Do not invent names, numbers or ROI; mark anything unbenchmarked as "needs pilot". Skip heavily regulated segments outside scope. Make it actionable within 7 days.
Claude

Flyer, reel or page to a structured card

Turn messy promotional content (flyers, reels, posts) into a clean, structured record with confidence scores.

Why it worksThe pattern behind my event-intelligence agent. It separates what is confirmed from what is guessed.

Show the prompt
Starter prompt · copy and adapt
You are an extraction analyst. Read the attached item (image, video transcript, caption or page) the way a sharp human researcher would.

1. Decide: is this an event (or [your record type])? Answer yes/no with one line of evidence.
2. Extract into JSON:
   name, date, start_time, timezone, venue, address, neighborhood, price, ticket_or_rsvp_link, lineup_or_speakers, audience_or_vibe, dress_code, source_url
3. For EVERY field add:
   - value (or null)
   - status: "confirmed" (quote the evidence) | "inferred" (explain) | "missing"
4. Give an overall confidence score from 0 to 1 and the reason.
5. List missing fields a human should check before publishing.

Never fill a field to look complete. A null with a reason beats a confident guess.
Any model

Inquiry reply drafter (you approve before it sends)

First replies to new customer inquiries, drafted from your approved facts only.

Why it worksThe best first automation is a draft, not an auto-send. This one asks for missing details instead of inventing them.

Show the prompt
Starter prompt · copy and adapt
You draft first replies to new customer inquiries for [business name], a [type of business] in [area].

APPROVED FACTS (use nothing else):
- Services: [list]
- Hours: [hours]
- Service area: [area]
- Price ranges you may share: [ranges, or "none, offer a call"]
- Booking link: [link]

For each inquiry:
1. Classify it: new job / quote request / existing customer / not a fit / urgent.
2. List the facts you used and the information that is missing (date, address, budget, scope).
3. Draft a reply under 90 words: warm, specific to their message, asks for the missing details, offers the booking link.
4. Flag anything that needs the owner: pricing outside approved ranges, complaints, legal or safety issues.

Never promise a price, date or outcome that is not in the approved facts.

INQUIRY:
[paste]
Any model

Compare three quotes without losing the details

Contractor, vendor or software quotes that use different language and line items.

Why it worksThe cheapest total is not the same as the most comparable offer. This keeps evidence attached to every number.

Show the prompt
Starter prompt · copy and adapt
Compare the attached quotes. Remove personal data you do not need first.

Build a comparison table with these fields for each quote:
total, labor, materials, quantities, brand/model, permits, timeline, warranty, payment schedule, exclusions, unknowns.

Rules:
- Cite the page or section for every value.
- Write "not stated" instead of guessing.
- Normalize carefully and show your math.

Then list:
1. Mismatched scope between quotes (quantities, allowances, disposal, permits).
2. Questions I should ask each vendor before signing.
3. What you cannot judge from paper (workmanship, insurance, licensing).

Do not choose the vendor. Organize the evidence so I can.
Any model

Messy meeting to five next actions

Right after a call, with the transcript or rough notes.

Why it worksThe win is not a prettier summary. It is a smaller gap between the conversation and the work.

Show the prompt
Starter prompt · copy and adapt
From the transcript or notes below, return ONLY a table with four columns:
action | owner | due date | open question

Rules:
- Only include commitments someone actually made. Mark suggestions as "suggested".
- If no date was said, write "no date set". Do not invent deadlines.
- Max 8 rows, most important first.
- Under the table, list decisions made (one line each) and anything that needs a decision-maker.

NOTES:
[paste]
Claude

The owner's morning brief (scheduled task)

Set it as a scheduled task with your calendar and email connected. Runs before your first appointment.

Why it worksTen minutes of sorting every morning, done before you wake up. My own day starts with one.

Show the prompt
Starter prompt · copy and adapt
Every weekday at 6:30am, prepare my morning brief from my connected calendar and inbox.

Sections, in this order, under 250 words total:
1. TODAY: appointments with time, place, and one line of prep for each.
2. NEEDS ME: emails that require a decision or reply from me today (max 5), each with a one-line suggested response.
3. WAITING ON OTHERS: threads where I am waiting, older than 3 days.
4. MONEY: invoices due or overdue mentioned in email this week.
5. ONE THING: the single most important task for today, and why.

Ignore newsletters and promotions. Do not send or reply to anything. If something looks urgent overnight, put it first with "URGENT".
Any model

Rambling voice note to a clean SOP

Record yourself explaining a task once. Get a procedure your team can follow.

Why it worksMost small-business know-how lives in one head. This is the cheapest way to get it out.

Show the prompt
Starter prompt · copy and adapt
Turn this voice-note transcript into a standard operating procedure.

Format:
- Title and purpose (one sentence)
- When to use it / trigger
- Who owns it
- Tools and access needed
- Steps, numbered, one action per step, in the order they happen
- Decision points: "If X, then Y"
- What good looks like (quality check)
- Common mistakes
- Questions for the owner: anything unclear or contradictory in the transcript

Keep my wording where it is specific. Do not add steps I did not describe; list suggested additions separately.

TRANSCRIPT:
[paste]
Any model

Weekly client status pack

A one-page weekly update built from a tracking sheet and your notes.

Why it worksEvery number traces back to a cell. Clients get a consistent update; you get an hour back each week.

Show the prompt
Starter prompt · copy and adapt
Build this week's one-page status update for [client] from the sheet and notes below.

Sections:
1. What changed this week (3 to 5 bullets)
2. Numbers vs last week: [metric list], with the change and a one-line why
3. Blockers and who owns each
4. Next three actions with dates

Rules:
- Every number must match a cell in the sheet. Cite the cell (e.g. B12).
- No claim without a source in the notes.
- Plain language. Under 300 words.
- Then draft a 5-line email to send it, which I will review before sending.

SHEET:
[paste or attach]
NOTES:
[paste]

Run the five-leak audit on your own business

Pair it with the free Bottleneck Audit framework. Talk through your week, get a ranked map.

Why it worksThis is the same structure as my free audit call, so you can do it yourself first.

Show the prompt
Starter prompt · copy and adapt
Act as an operations consultant running a 45-minute AI audit on my business. Interview me, one question at a time.

Cover five areas where hours leak out of a business:
1. Hours: the owner's time
2. Bookings: inquiries, follow-up, missed calls
3. Repeat work: the same tasks every day or week
4. Numbers: reports and knowing where you stand
5. Know-how: what lives in one person's head

Also ask which systems I use and where my customer information lives.

When we are done, give me a one-page map:
- What you heard (3 lines, my words)
- The top 3 hour leaks: job, hours per week today, what it looks like fixed
- What should stay human, and why
- Path A (one targeted fix first) or Path B (a rollout, one team at a time), with your reasoning
- What to measure for the next 60 days

Name hours, not tools. Do not recommend software until the problem is pinned down.

New prompts, every week.

The newest prompt in the library and the real use behind it. One email a week. Unsubscribe anytime.

One email a week at most. Unsubscribe anytime.

A prompt is a start. A workflow is a result.

Bring the job to the free audit and leave with a map.

Book your free AI audit