Use: at the top of nearly every other template. The single highest-leverage thing on this page.

MY CONTEXT (reuse everywhere):
Business: [WHAT YOU SELL, ONE PLAIN SENTENCE]
Stage/size: [REVENUE, HEADCOUNT]
Customer: [WHO BUYS — role, company size, or life situation]
Their problem: [THE PAIN THEY PAY TO REMOVE, IN THEIR WORDS]
We replace: [WHAT THEY DO INSTEAD TODAY]
Proof: [2-3 SPECIFIC RESULTS WITH NUMBERS OR NAMES]
Only true of us: [REAL DIFFERENTIATOR — or say you don't have one]
Constraints: [BUDGET, TIME, HEADCOUNT, RUNWAY]
Already tried: [SO YOU DON'T GET RECYCLED SUGGESTIONS]

STANDING RULES:
- Never invent a fact, statistic, or capability. If you need
  something I haven't given you, ask.
- Flag anything you're inferring as [ASSUMPTION].
- Be direct. No preamble, no encouragement.

Use: before any significant decision. Inverts the model’s bias toward agreeing with you.

[CONTEXT BLOCK]
I'm about to [DECISION].
It's 12 months from now and this failed clearly.
1. Write the post-mortem — what went wrong, ranked by
   likelihood. Specific to my business, not generic advice.
2. For each: the EARLIEST signal I could have caught it by.
3. Which failures are fatal vs recoverable?
4. What would I have to believe for this to work? Mark
   which of those are untested.
Do not reassure me. Assume failure and work backwards.

Use: when growth has stalled and you’re not sure why.

[CONTEXT BLOCK]
My numbers, last [PERIOD]: [PASTE METRICS]
I think my bottleneck is [YOUR HYPOTHESIS].
1. What do these numbers suggest the real constraint is?
   Argue against my hypothesis if they don't support it.
2. Which single number, improved 20%, most changes my
   outcome? Show the reasoning.
3. What data am I missing that would change your answer?
4. The smallest test to confirm the real constraint in
   2 weeks.

Use: when a choice keeps getting reopened. Forces the trade-off into the open.

[CONTEXT BLOCK]
Decision: [WHAT I'M CHOOSING BETWEEN]
Deadline: [WHEN] · Reversible? [YES/NO/PARTLY]
Structure this for me:
1. What am I ACTUALLY choosing between? State it in one
   sentence per option, stripped of framing.
2. What does each option cost that the other doesn't —
   money, time, optionality, morale?
3. What would have to be true for each to be right?
4. What information would change my mind, and can I get
   it before the deadline?
5. If reversible, what's the cheapest version I could
   try first?
6. Name the option I appear to have already chosen
   emotionally, and what's driving that.

Use: turning a vague idea into something you can actually test.

[CONTEXT BLOCK]
My idea: [2-4 PLAIN SENTENCES]
Who it's for: [SPECIFIC — not "small businesses"]
Don't evaluate whether it's good. Instead:
1. Rewrite as a testable problem statement: "[PEOPLE]
   experience [PAIN] when [SITUATION] and cope by
   [WORKAROUND]."
2. List every assumption that must be TRUE, split into:
   problem / customer / solution / money / channel.
3. Mark each TESTED or UNTESTED, and FATAL or SURVIVABLE
   if false.
4. Rank fatal+untested by cheapest to test. That's my
   to-do list.
5. Name the assumption I seem most attached to.

Use: content planning. Starts from real customer questions, not keyword lists.

[CONTEXT BLOCK]
Generate 25 content topics from real customer questions.
Inputs:
- Objections in sales calls: [LIST 3-5]
- Questions asked before buying: [LIST 3-5]
- Mistakes I watch prospects make: [LIST 3-5]
Return a table: | question as a customer would type it |
intent | the angle only WE can take | format | buying stage |
Rules: every title is a question or specific claim, never
a vague noun phrase. Skip anything a competitor could
publish identically. Flag topics where we lack the proof
to be credible.

Use: structuring an article to be quoted by AI engines, not just ranked.

[CONTEXT BLOCK]
Topic: [THE QUESTION THIS ANSWERS]
Build an outline optimised to be QUOTED, not just ranked:
1. A one-sentence direct answer, under 40 words, written
   so it can be lifted verbatim.
2. 3-5 key takeaways as a scannable list.
3. Headings phrased as questions people actually ask.
4. Under each, a self-contained 40-120 word passage that
   makes sense pulled out of context.
5. At least one comparison table.
6. 5-8 FAQ questions targeting long-tail variations.
For each section: the specific claim it makes and the
evidence supporting it. Flag any section where we'd be
asserting something without proof.

Use: repurposing. Persona is fine here — style is the job.

[CONTEXT BLOCK]
Voice sample to match: "[PASTE 2 PARAGRAPHS OF YOUR
BEST WRITING]"
Source material: [ARTICLE / CUSTOMER STORY / RESULT /
POINT OF VIEW]
Turn this into platform-native posts — genuinely different
expressions, not the same text reformatted:
LINKEDIN: 120-180 words. Open with a specific claim or
  concrete moment, never a question.
X: under 280 characters. Sharpest version of the claim.
INSTAGRAM: caption, first line works as a standalone hook.
NEWSLETTER: 50 words that earn a click without clickbait.
For each: 3 hook variations.
Banned: "Here's the thing", "Let that sink in", engagement
bait, and any post that would work for a competitor.

Use: when a draft is fine but the opening is flat.

[CONTEXT BLOCK]
My draft: [PASTE]
1. Rate the first line — would someone stop scrolling? If
   no, say which word lost them.
2. Rewrite the opening 5 ways: a specific number; a
   contrarian claim we can defend; a concrete moment; a
   named mistake the reader is making; the plainest
   statement of the point.
3. For each, the risk (too aggressive, vague, overpromises).
4. Which one does the BODY actually deliver on? A hook the
   piece doesn't earn costs more than a boring one.

Use: when everything you write sounds like everyone else.

[CONTEXT BLOCK]
Our current pitch: "[HEADLINE / ELEVATOR PITCH]"
Competitors say: [COMPETITOR 1] · [2] · [3]
1. Strip out every claim a competitor could also make.
   What's left that's genuinely only true of us?
2. If nothing is left, say so plainly — that's the finding.
3. Rewrite around what survived. 3 options, under 15 words,
   in customer language not marketing language.
4. For each: the customer it wins and the customer it
   loses. Trade-offs are the point.

Use: before outreach. Run on a model with live web access.

[CONTEXT BLOCK]
Prospect: [NAME], [TITLE] at [COMPANY] ([URL])
Research them, then give me:
1. THREE specific, recent, verifiable company facts
   (last 6 months) — funding, hires, launches, press.
2. TWO signals they may have the problem I solve, with
   the evidence for each.
3. ONE thing about this person — post, talk, career move.
4. The single most likely reason they'd ignore my email.
Rules: only facts you can point to a source for. If you
can't verify something, write NOT FOUND rather than
guessing. Rank by relevance to the problem I solve.

Use: pipeline review. Built to disqualify, not to reassure.

[CONTEXT BLOCK]
Why we usually lose: [TOP 3 LOSS REASONS]
The deal: [COMPANY, CONTACT + WHETHER THEY HOLD BUDGET,
STAGE, WEEKS IN PIPELINE]
What they SAID: [QUOTES]
What they DID: [ACTIONS WITH A COST — or "nothing"]
Next step booked: [YES + DATE / NO]
Find reasons this WON'T close.
1. Split SAID from DID. Only DID predicts anything.
2. For each — pain, power, urgency, process, competition —
   mark KNOWN / ASSUMED / UNKNOWN.
3. Which of my loss patterns does this most resemble?
4. Score LIKELY / AT RISK / LIKELY DEAD. Treat UNKNOWN as
   negative, not neutral.
5. The one question that would most change this. Exact
   wording.

Use: once per offer. Build it before the call, not during.

[CONTEXT BLOCK]
Act as a skeptical qualified buyer: [TITLE] at a
[COMPANY TYPE]. You have budget, the problem is real,
you're unconvinced.
Our offer: [WHAT AND PRICE]
1. The 10 objections you'd actually raise, ranked by how
   likely each is to kill the deal.
2. For each: the STATED objection, the REAL concern
   underneath, and the evidence that would resolve it.
3. A response to each under 60 words — acknowledge without
   agreeing, one clarifying question, specific evidence,
   concrete next step.
4. Flag which objections are NOT sales problems — where
   our product or price is genuinely wrong for them.
5. For the top 3: the question I could ask EARLIER so the
   objection never forms.

Use: within an hour of any call.

[CONTEXT BLOCK]
My raw call notes: [PASTE — messy is fine, include their
exact words where you have them]
Write the follow-up:
- Open by referencing something SPECIFIC they said, in
  their words not mine.
- Confirm the problem as THEY framed it.
- One piece of evidence for their specific concern.
- The next step as agreed, with a date.
Under 120 words. No "great speaking today", "as discussed",
"circling back". One CTA. Only use facts in my notes — if
you need something I didn't record, ask. Do NOT invent a
detail about the conversation.
Then separately: 3 things I should have asked but didn't,
and whether it's still possible to ask them.

Use: before replying to anything non-trivial. Catches churn signals.

[CONTEXT BLOCK]
Known current issues: [OUTAGES, DELAYS, RECENT PRICE OR
POLICY CHANGES]
The ticket: [PASTE VERBATIM]
About this customer: [TENURE, SPEND, PAST TICKETS, RECENT
ACCOUNT EVENTS — or "nothing"]
Diagnose. Don't draft a reply yet.
1. Literal request. 2. Underlying problem — what likely
happened just before they wrote in?
3. Emotional state 1-5, quoting the words that tell you.
4. CHURN SIGNAL — export requests, contract/notice queries,
   competitor mentions, cooling tone vs previous tickets,
   timing near a price change. Rate LOW/MED/HIGH.
5. The real question they haven't asked.
6. Route: factual answer, or needs a human?
7. What I'd need to know to handle this properly.

Use: when you got it wrong. A handled failure can produce more loyalty than none.

[CONTEXT BLOCK]
What went wrong: [HONESTLY, INCLUDING OUR PART]
Impact on them: [WHAT IT COST THEM]
What I can offer: [REAL OPTIONS AND LIMITS]
Write the recovery message:
1. Own it in the first sentence. Active voice — "we sent
   it to the wrong address", not "an error occurred".
2. Show we understand the actual impact on them.
3. State the remedy concretely, and when. Offer it; don't
   make them ask.
4. What changes so it doesn't recur — only if true.
5. No excuses, no internal explanations.
Under 130 words. Apologise once, early, specifically.
Ask them for nothing. Don't upsell.
Then flag: is the remedy proportionate to the harm?

Use: on cancellation. Diagnose the cause before offering anything.

[CONTEXT BLOCK]
What they said: [CANCELLATION MESSAGE, VERBATIM]
History: [TENURE, SPEND, USAGE TREND, PAST TICKETS]
What I can offer: [REAL OPTIONS]
1. Which cause is it? (A) price (B) unmet expectation
   (C) specific failure (D) their circumstances changed.
   State confidence and quote the evidence.
2. Is this saveable? If it's D, say so — a save attempt
   wastes their time and burns the relationship.
3. The right response for THIS cause. Don't offer a
   discount unless the cause is genuinely price.
4. Draft it, under 120 words. Don't guilt them. Make
   cancelling easy regardless.
5. What should we log about product or onboarding?

Use: documenting a process only you know. The one-question-at-a-time rule is essential.

[CONTEXT BLOCK]
I need to document how I do [PROCESS] so someone else
can run it without me.
INTERVIEW ME. Ask ONE question at a time and wait for my
answer. Don't produce the SOP until you have enough.
Start by asking me to walk through the LAST TIME I did it.
Probe for: steps I do automatically and would forget to
mention; decisions I make and what I base them on; what
"done well" means vs merely "done"; what goes wrong and
how I spot it; what I check before finishing.
Every time I say "it depends" — STOP and ask exactly what
it depends on until the rule is explicit.
Then produce: purpose, trigger, steps, decision rules,
quality checks, common failures, escalation criteria.
Finally, list every question you still have.

Use: quarterly, or when you’re the bottleneck.

[CONTEXT BLOCK]
Everything I personally did last week: [LIST WITH ROUGH
HOURS]
Sort every task:
A) Only the founder can do this (relationships, judgment,
   final decisions)
B) Could be delegated to a person today
C) Could be systematised or automated
D) Shouldn't be done at all
For B: the cheapest role that could own it.
For C: describe the system in one sentence.
For D: what happens if it stops.
Then: what % of my week is in A? If under 40%, which single
change moves the needle most?

Use: before writing any job ad. Most bad hires trace to a vague scorecard.

[CONTEXT BLOCK]
Role: [TITLE] · Why now: [WHAT'S BREAKING]
Budget: [RANGE] · Reports to: [WHO]
Before any job description, define success:
1. 4-6 OUTCOMES they must deliver in 12 months. Measurable,
   specific to my business, not generic responsibilities.
2. For each, the ONE competency that predicts it.
3. What evidence in someone's history demonstrates it.
4. A written 1-3-5 rubric per competency — what a weak,
   adequate and strong answer contains.
5. Must-have vs nice-to-have. Challenge anything I marked
   must-have that probably isn't.
6. Flag if this is actually TWO roles.
7. Flag any requirement that's a proxy for background
   rather than ability — degrees, "X+ years", "culture fit".

Use: testing an SOP before a real person follows it.

The document: [PASTE YOUR SOP]
You're a new starter on day three. Competent, motivated,
but you know NOTHING about this business beyond what's
written here. You can't ask anyone.
Follow this document literally, step by step. At every
step tell me: what you'd do; what you'd have to GUESS;
where you'd get STUCK; where you'd confidently do the
WRONG thing because it's ambiguous but sounds clear; any
term you don't understand.
Be literal to the point of pedantry. Do NOT use common
sense to fill gaps — common sense is exactly what a new
starter lacks about MY business.
Then: critical gaps (wrong result, unnoticed) · friction
gaps · assumed knowledge · the step most likely done wrong
· the 3 questions you'd ask on day one.

Use: monthly. Formulas only — model accuracy on financial calculation is unreliable.

[CONTEXT BLOCK]
Current period: [REVENUE, COST OF SALES, EACH OVERHEAD,
NET PROFIT]
Comparison period: [SAME LINES, 6-12 MONTHS EARLIER]
DO NOT CALCULATE. Give me formulas and tell me what to
compute, then interpret what I report back.
1. Each metric with its formula — gross margin %, net
   margin %, each overhead as % of revenue, period change.
2. Which lines moved most in absolute terms, and which as
   a proportion of revenue? Different questions, both matter.
3. If gross margin moved: what would current revenue have
   produced at the OLD margin? That gap is usually the story.
4. Largest overhead — what's a 10% cut worth annually?
5. What this P&L does NOT tell me before I draw conclusions.
6. Three questions to ask my accountant.

Use: get a model you can edit, not a number you can’t check.

[CONTEXT BLOCK]
Cash today: [AMOUNT] · Owed to me: [AMOUNTS + DATES]
I owe: [AMOUNTS + DATES — include tax/VAT]
Monthly revenue, last 6: [FIGURES]
Monthly costs, fixed vs variable: [ITEMISED]
Payment terms: [WHEN CUSTOMERS ACTUALLY PAY]
Build a 12-month cash flow as a SPREADSHEET STRUCTURE.
Do not calculate.
1. Exact rows/columns, with an ASSUMPTIONS block at top and
   calculation rows referencing it. Never hard-code a number
   inside a formula.
2. Formulas cell by cell — cash received (timing-adjusted,
   NOT revenue billed), cash out, closing cash, runway.
3. Every assumption labelled, flagging which you inferred.
4. Cash items founders forget — tax, annual renewals,
   seasonality. Ask me whether each applies.
5. The 3 inputs that most change the outcome if wrong by 20%.
Remember: profit ≠ cash. Model when money MOVES.

Use: on anything important before it ships. Never ask “is this good?”

Act as [THE SPECIFIC PERSON WHO'LL RECEIVE THIS — a busy
CFO, a skeptical prospect, a journalist looking for holes].
You see a lot of these and dismiss most in seconds.
Here's what you just received: [PASTE]
1. YOUR HONEST REACTION — the actual thought as you read
   the first line. Be blunt. If you'd stop reading, say
   which word lost you.
2. THE THREE OBJECTIONS you'd raise, ranked by likelihood.
3. THE REWRITE — rewrite it so it survives your own
   objections. Change only what's necessary.
Then step out of character: what single change would most
improve the odds of this working?

Use: whenever a prompt works well. This is how the library grows.

A prompt I wrote produced good output. Turn it into a
reusable template.
The prompt: [PASTE]
The output it gave: [PASTE]
What I'd want next time: [SAME THINKING, DIFFERENT
SITUATION — describe one]
1. SEPARATE — which parts describe this specific situation,
   and which describe the thinking I want repeated? Only
   the second belongs in a template.
2. VARIABLES — replace every situation-specific detail with
   [DESCRIPTIVE NAME]. Names must be clear enough that a
   colleague knows what to supply.
3. MAKE IMPLICIT THINGS EXPLICIT — what did this prompt get
   away with because of context that won't exist next time?
   Add: output format, length, banned phrases.
4. ADD UNCERTAINTY HANDLING — flag inferences, NOT FOUND
   rather than guessing.
5. THE TEMPLATE — write it out.
6. STRESS TEST — run your template against the different
   situation I described. If it only works for my original
   case, it's still a prompt. Fix it and say what you changed.