Choose one decision.
Supply the bracketed inputs, run the prompt, review the evidence, and choose a real next action. Every copy button includes the session instructions. These prompts do not publish, contact customers, or execute campaigns.
Independently authored. HubSpot’s public staged prompt-resource concept informed the request; its gated prompts were not accessed or copied.
40 of 40 prompts
WM-D01 · Discover
Find the problem in real conversations
Bring: [anonymized conversations], [business idea]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [anonymized conversations], [business idea]. Task: Identify specific situations people described, their workarounds, and consequences. Separate enthusiasm about an idea from evidence of an important problem. Return: A problem map with source notes and the three most useful follow-up questions. Before finalizing: Can you point to an actual observation behind each proposed problem?
WM-D02 · Discover
Prepare a first customer conversation
Bring: [person you can reach], [uncertain assumption]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [person you can reach], [uncertain assumption]. Task: Write neutral questions about the last time the situation occurred. Explore what the person tried and how they decided. Avoid pitching before understanding. Return: A ten-minute conversation guide and a simple evidence note. Before finalizing: Remove questions that invite politeness rather than useful information.
WM-D03 · Discover
Choose an audience you can reach
Bring: [skills], [audience access], [candidate problems], [time]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [skills], [audience access], [candidate problems], [time]. Task: Compare plausible audiences by problem relevance, ability to help, and access to real conversations. Keep market size and willingness to pay unknown unless evidenced. Return: Three audience options with one practical research step each. Before finalizing: Prefer reachable learning over an invented large opportunity score.
WM-D04 · Discover
Understand the current workaround
Bring: [customer account], [task steps], [known tools]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [customer account], [task steps], [known tools]. Task: Describe how the person solves the problem today, including doing nothing. Identify effort, switching friction, and what already works well. Return: A before-state map and questions about why a change would be worth it. Before finalizing: Do not assume an imperfect workaround means someone will buy a replacement.
WM-D05 · Discover
Compare alternatives without guessing
Bring: [alternative pages], [dates], [your proposed offer]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [alternative pages], [dates], [your proposed offer]. Task: Compare visible promises, intended users, scope, and evidence. Mark missing information as unknown. Include manual work and avoiding the problem as alternatives. Return: A compact comparison and two differentiation hypotheses to ask buyers about. Before finalizing: Do not equate absent website copy with absent functionality.
WM-D06 · Discover
Check your useful advantage
Bring: [experience], [work samples], [offer idea], [buyer problem]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [experience], [work samples], [offer idea], [buyer problem]. Task: Identify advantages you can demonstrate today and capabilities still needing development. Suggest proof you could produce honestly. Return: An advantage statement, evidence list, and one proof-building task. Before finalizing: Exclude invented credentials, customer outcomes, and superiority claims.
WM-D07 · Discover
Write a voice guide that sounds like you
Bring: [your approved writing], [audience], [tone preferences]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [your approved writing], [audience], [tone preferences]. Task: Turn your natural voice into clear instructions for future drafts. Include vocabulary, sentence style, warmth, directness, and claim boundaries. Return: A short voice card with three before-and-after examples. Before finalizing: Read it aloud and remove language you would never actually say.
WM-D08 · Discover
Make your story understandable
Bring: [origin], [customer problem], [offer], [verified proof]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [origin], [customer problem], [offer], [verified proof]. Task: Draft a concise explanation linking why you care with the help you provide. Keep the customer’s need visible and avoid turning aspirations into results. Return: A one-sentence promise, a short story, and a comprehension question. Before finalizing: Would a new reader know what you help with and what remains unproven?
WM-D09 · Discover
Map the first useful result
Bring: [offer], [customer starting point], [delivery steps]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [offer], [customer starting point], [delivery steps]. Task: Define the smallest outcome that demonstrates value after purchase or signup. Separate the customer’s benefit from administrative activity. Return: A first-value statement, required steps, likely friction, and observation method. Before finalizing: A login, download, or completed profile alone may not be a useful result.
WM-D10 · Discover
Choose the next research move
Bring: [known facts], [assumptions], [available time], [decision]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [known facts], [assumptions], [available time], [decision]. Task: Rank uncertainties by whether they would change your next commitment. Suggest the smallest real-world investigation for the most important one. Return: One research task with input, method, evidence, and review date. Before finalizing: Do not substitute simulated customers for actual demand evidence.
WM-C01 · Decide
Shape a small pilot offer
Bring: [problem evidence], [skills], [capacity], [costs]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [problem evidence], [skills], [capacity], [costs]. Task: Define a bounded service or product test with a clear deliverable, exclusions, customer effort, and delivery timing. Identify what must be confirmed before charging. Return: A pilot offer card and three unresolved questions. Before finalizing: Can you deliver the stated outcome within the time and resources available?
WM-C02 · Decide
Choose your first three audience groups
Bring: [declared goals], [available fields], [offer]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [declared goals], [available fields], [offer]. Task: Suggest no more than three groups that need different help. Use reliable facts and stated preferences. Define an appropriate general path for unknown records. Return: Group purpose, simple rule, next action, and exit condition. Before finalizing: Do not infer readiness or identity from a single page visit.
WM-C03 · Decide
Check who should receive the message
Bring: [segment rule], [permission fields], [purchase status]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [segment rule], [permission fields], [purchase status]. Task: Review inclusion, exclusions, overlap, and timing. Test a qualifying person, an opt-out, a customer, and someone with missing information. Return: A plain-language rule and edge-case decisions. Before finalizing: Keep permission checks outside every OR branch and recheck before sending.
WM-C04 · Decide
Adapt one offer for different needs
Bring: [approved offer], [audience groups], [voice card]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [approved offer], [audience groups], [voice card]. Task: Explain the same offer differently for explorers, evaluators, or customers where appropriate. Change the guidance and examples without changing factual terms. Return: Three short messages with a suitable next step for each. Before finalizing: Remove redundant purchase invitations for people who already bought.
WM-C05 · Decide
Create a useful free first step
Bring: [customer problem], [paid offer], [delivery capacity]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [customer problem], [paid offer], [delivery capacity]. Task: Design a small resource that helps someone make immediate progress and naturally relates to your work. Specify what it does and does not solve. Return: A resource outline, completed-example idea, and optional next action. Before finalizing: Is it useful without forcing a sales call or pretending to be a complete solution?
WM-C06 · Decide
Plan a truthful offer page
Bring: [offer card], [example], [questions], [approved proof]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [offer card], [example], [questions], [approved proof]. Task: Outline a page that explains fit, value, scope, effort, price context, delivery, and next steps. Flag missing facts rather than filling them in. Return: Page sections, draft key copy, and a fact-check list. Before finalizing: Do not invent testimonials, scarcity, guarantees, or completed client work.
WM-C07 · Decide
Give new subscribers a small win
Bring: [signup promise], [resource], [email permissions]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [signup promise], [resource], [email permissions]. Task: Plan a short welcome sequence that delivers the resource and helps the reader apply one step. Make timing and stop rules explicit. Return: Three message outlines with purpose, action, and exit criteria. Before finalizing: Subscription must be voluntary and delivery must work before promotion.
WM-C08 · Decide
Welcome the person who bought
Bring: [purchase], [first-value milestone], [support options]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [purchase], [first-value milestone], [support options]. Task: Create a clear starting experience for a new buyer. Explain what happens now, what they must do, and how to get help. Return: Confirmation copy, first-step guidance, and a follow-up check. Before finalizing: Separate necessary purchase communication from optional promotional email.
WM-C09 · Decide
Choose a simple website variation
Bring: [visitor-selected goal], [base content], [available implementation]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [visitor-selected goal], [base content], [available implementation]. Task: Propose a useful content variation based on an explicit choice, with a general fallback. Avoid assuming the site knows the person’s identity. Return: Choice labels, two content variants, fallback, and success measure. Before finalizing: Keep the experience useful without tracking or a known contact record.
WM-C10 · Decide
Make the offer affordable to deliver
Bring: [price], [delivery time], [variable costs], [capacity]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [price], [delivery time], [variable costs], [capacity]. Task: Estimate contribution and capacity under clearly stated assumptions. Show what would make the offer unsuitable and what you need to measure during the pilot. Return: A simple economics table and a delivery limit to review. Before finalizing: Do not present illustrative income as a forecast or ignore your own work time.
WM-S01 · Share
Plan a demonstration people can use
Bring: [offer], [common problem], [sample inputs]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [offer], [common problem], [sample inputs]. Task: Outline a short demonstration that creates value on its own. Use fictional sample data where needed and label it clearly. Connect the lesson to a relevant optional next step. Return: A demonstration script, required assets, and a claim check. Before finalizing: Do not show fabricated client results as evidence.
WM-S02 · Share
Pick one place to start
Bring: [audience behavior], [your access], [time], [content skills]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [audience behavior], [your access], [time], [content skills]. Task: Compare a small number of realistic channels. Choose one based on audience fit and your ability to contribute consistently. Return: A first-channel decision and a two-week participation plan. Before finalizing: Do not promise reach or sales from an untested channel choice.
WM-S03 · Share
Contribute to a community appropriately
Bring: [community rules], [actual question], [your expertise]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [community rules], [actual question], [your expertise]. Task: Draft a complete useful answer and identify whether linking a resource is permitted. Offer a nonpromotional version when a link is not appropriate. Return: An answer draft and a participation note. Before finalizing: Do not invent conversations or treat group membership as permission for mass messages.
WM-S04 · Share
Propose a small collaboration
Bring: [potential collaborator], [verified overlap], [useful topic]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [potential collaborator], [verified overlap], [useful topic]. Task: Suggest a joint resource or session that serves both audiences. Draft a specific invitation and list the details that need agreement. Return: One collaboration concept, short draft, and responsibilities checklist. Before finalizing: No outreach is sent; audience size, prices, and agreement remain unverified unless supplied.
WM-S05 · Share
Turn one lesson into three posts
Bring: [approved lesson], [channel], [voice card]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [approved lesson], [channel], [voice card]. Task: Create an explanation, a demonstration, and an objection answer. Give each standalone value and keep the next step relevant. Return: Three distinct drafts with factual claims listed for review. Before finalizing: Do not just change the opening line while repeating the same post.
WM-S06 · Share
Write a useful newsletter edition
Bring: [promised newsletter topic], [lesson], [reader stage]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [promised newsletter topic], [lesson], [reader stage]. Task: Draft one edition that teaches a concrete step, includes a simple example, and offers one optional action. Keep it aligned with the signup promise. Return: Subject options, email body, and one CTA. Before finalizing: Remove exaggerated urgency and unverified results.
WM-S07 · Share
Choose a search question to answer
Bring: [customer questions], [existing pages], [expertise]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [customer questions], [existing pages], [expertise]. Task: Select a narrow question you can answer usefully. Outline an original page with examples, limitations, and relevant links. Identify overlap with existing content. Return: One article brief and a plan for showing expertise. Before finalizing: Do not invent keyword volume or guarantee rankings.
WM-S08 · Share
Prepare an AI-discovery evidence check
Bring: [buyer questions], [public pages], [available search tools]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [buyer questions], [public pages], [available search tools]. Task: Design a small review of actual AI answers and citations. Define what to record and how to distinguish missing coverage from inconsistent outputs. Return: A query sample, observation sheet, and content-improvement questions. Before finalizing: Do not say a scan happened unless actual outputs were supplied or retrieved.
WM-S09 · Share
Plan a first paid-traffic test
Bring: [offer economics], [landing page], [keyword or creative ideas], [loss limit]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [offer economics], [landing page], [keyword or creative ideas], [loss limit]. Task: Choose one testable paid-traffic hypothesis. Define the audience, action, budget limit, monitoring, and evidence needed before scaling. Return: A test card and readiness checklist. Before finalizing: Paid traffic does not validate an unclear offer; no ad is launched by this prompt.
WM-S10 · Share
Build a calendar you can finish
Bring: [available hours], [approved assets], [review needs], [dates]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [available hours], [approved assets], [review needs], [dates]. Task: Schedule a small amount of production, distribution, and response time. Leave room for learning and delivery. Label dates as proposed. Return: A two-week calendar with one clear purpose per activity. Before finalizing: Do not fill every day simply to appear consistent or schedule beyond capacity.
WM-I01 · Improve
Define evidence before the test
Bring: [business question], [action], [observation window]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [business question], [action], [observation window]. Task: Write what you expect to learn, what you will record, and what different results would mean. Include one guardrail and an inconclusive outcome. Return: An experiment card with goal, action, evidence, checkpoint, and next-move options. Before finalizing: Do not change success criteria after seeing the outcome without saying so.
WM-I02 · Improve
Make sense of the actual response
Bring: [test record], [observed results], [cost and time]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [test record], [observed results], [cost and time]. Task: Separate what happened from explanations. Suggest two plausible reasons and the cheapest useful next check. Preserve small-sample uncertainty. Return: A short evidence review and one recommended next move. Before finalizing: Likes, downloads, requests, purchases, and delivered value are different signals.
WM-I03 · Improve
Find the first customer-journey obstacle
Bring: [visits], [actions], [follow-up], [customer feedback]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [visits], [actions], [follow-up], [customer feedback]. Task: Identify where progress appears to stop and which data is missing. Distinguish a broken step from weak interest or delayed decision-making. Return: A bottleneck hypothesis and a check you can perform today. Before finalizing: Do not prescribe more traffic before checking the current journey.
WM-I04 · Improve
Review your audience groups
Bring: [group rules], [current examples], [outcomes], [overlap]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [group rules], [current examples], [outcomes], [overlap]. Task: Check whether the groups still describe meaningful needs and produce useful differences. Suggest a merge, revision, or retirement where appropriate. Return: A group-by-group review with one simplification. Before finalizing: Do not keep a segment simply because creating it took effort.
WM-I05 · Improve
Ask buyers about the first result
Bring: [promised outcome], [delivery experience], [contact permission]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [promised outcome], [delivery experience], [contact permission]. Task: Prepare neutral questions about what the buyer used, what helped, and what was confusing. Ask about observed experience rather than general satisfaction alone. Return: A short feedback request and a note-taking template. Before finalizing: Do not turn feedback into a public testimonial without separate permission.
WM-I06 · Improve
Review the money and the time
Bring: [receipts], [spend], [delivery hours], [refunds], [unknown costs]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [receipts], [spend], [delivery hours], [refunds], [unknown costs]. Task: Separate revenue, variable contribution, overhead, and your work time. Calculate only what the inputs support. Identify what remains provisional. Return: A simple outcome summary and the next cost to verify. Before finalizing: Never treat gross sales as profit or missing expenses as zero.
WM-I07 · Improve
Improve a confusing explanation
Bring: [current copy], [actual misunderstanding], [offer facts]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [current copy], [actual misunderstanding], [offer facts]. Task: Revise the section that caused confusion while preserving the true scope. Explain the edit and how to test whether it helps. Return: Revised copy and one comprehension question. Before finalizing: Do not add a bigger promise to compensate for unclear wording.
WM-I08 · Improve
Decide whether to automate
Bring: [manual task], [frequency], [error examples], [time cost]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [manual task], [frequency], [error examples], [time cost]. Task: Assess whether the process is stable enough to automate. Define trigger, inputs, allowed action, review point, failure handling, and manual fallback. Return: An automation candidate brief or a reason to keep it manual. Before finalizing: A draft workflow is not a connected, tested automation.
WM-I09 · Improve
Capture a useful checkpoint
Bring: [goal], [completed work], [evidence], [blockers]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [goal], [completed work], [evidence], [blockers]. Task: Write a concise record that lets your future self resume. Link artifacts, keep unknowns visible, and select the smallest useful next action. Return: A checkpoint with decision, rationale, files, open questions, and next move. Before finalizing: Do not mark a task complete merely because an AI produced a draft.
WM-I10 · Improve
Choose continue, change, or stop
Bring: [experiment history], [constraints], [current evidence]
Use only supplied evidence or sources you actually retrieve. Separate observations, interpretations, and hypotheses. If a required input is missing, ask for it or state the limitation. Do not invent customer quotes, data, results, source access, or completed actions. Use minimized, appropriate data; do not infer sensitive traits. Produce a reviewable draft, not an external action. Context and inputs: [experiment history], [constraints], [current evidence]. Task: Compare continuing, changing one element, pausing, and stopping. Explain the tradeoffs and what new evidence would change the recommendation. Return: A decision note with a next action and review condition. Before finalizing: Do not pressure the founder to persist when delivery, economics, or circumstances do not support it.