Skip to main content
idapt
AccueilModèles IATarifs
Sign inStart free

Prompt library

Prompts that hold up at work

120 copy-paste prompts across 10 professions, each with the variables to fill, the reasoning that makes it work, and a one-click run in idapt.

Research

12 prompts

Outline a literature review

Organizing by schools of thought instead of paper-by-paper forces synthesis, which is what a review is graded on. The explicit no-invented-citations instruction converts the model's uncertainty into visible 'verify' markers rather than plausible fake references.

PlanOpen prompt

Summarize a research paper

The fixed section list stops the model from paraphrasing the abstract, and the admitted-versus-unadmitted limitations split is where the real critical reading happens. Requiring direct quotes for numbers makes hallucinated statistics easy to spot.

SummarizeOpen prompt

Extract findings from a study

A column schema with a location column makes every extracted number auditable against the source. The final abstract-versus-results check catches the most common spin in published work: conclusions that outrun the data.

ExtractOpen prompt

Generate research questions worth asking

The descriptive-causal-interventional ladder is how research methodologists actually taxonomize questions, so the output maps to study designs instead of vague curiosities. Forcing a confound, an obstacle, and a final ranked pick turns brainstorming into triage.

BrainstormOpen prompt

Draft unbiased survey questions

Making the model audit its own draft for leading and double-barreled questions catches most survey-design errors in one pass, and showing the rewrite teaches you the pattern. The left-out-questions section forces scope discipline, which is what keeps completion rates up.

WriteOpen prompt

Build a user interview guide

Behavior-first questions ('walk me through the last time') produce evidence; hypothetical questions produce polite fiction. Baking follow-up probes and listening signals into the guide keeps a live interview on track when answers go sideways.

PlanOpen prompt

Synthesize user interview notes

Requiring two sources per theme and quoting verbatims keeps the synthesis grounded in what people said rather than what the model expects users to say. Keeping contradictions visible instead of averaged is what makes the document trustworthy in a debate.

AnalyzeOpen prompt

Tear down a competitor's product

The observed-inferred-speculation labeling is the difference between competitive intelligence and fan fiction. Asking for their core bet first anchors the analysis in strategy rather than a feature checklist, which is where teardown documents usually go shallow.

AnalyzeOpen prompt

Outline a defensible market sizing

Two independent estimation paths that must roughly agree is the standard defense against garbage-in sizing, and explicit low-base-high ranges surface uncertainty instead of hiding it in one confident number. The labeled-placeholder rule keeps invented statistics out of your deck.

PlanOpen prompt

Check a source's credibility

Most citation failures are fit failures: the source is real but supports a weaker claim than the sentence it decorates. Separating provenance, method, and fit makes that gap explicit, and the ready-to-paste caveat sentence makes the safe path the easy path.

AnalyzeOpen prompt

Turn sources into an annotated bibliography

The three-move annotation structure (summary, evaluation, application) is the format instructors and reviewers actually expect, not a paragraph of vague praise. The [MISSING] convention stops the model from fabricating citation details, the classic failure of AI bibliographies.

WriteOpen prompt

Plan a focused research week

One falsifiable question per week is the difference between research and reading. The Wednesday kill criterion and the mandated slack come from project-management practice: plans without an exit condition and buffer quietly turn into three-week weeks.

PlanOpen prompt

Writing

12 prompts

Outline a blog post that earns its length

Deciding the reader's after-state and prior belief before outlining is what separates an argument from a topic dump. Requiring one claim, one proof, and one lived detail per section systematically eliminates the padding that makes AI-drafted posts unreadable.

PlanOpen prompt

Rewrite text for clarity

The preserve-facts-exactly instruction is the guardrail that makes AI editing safe on real documents: the model may not 'improve' your numbers or promises. The before-and-after offense list turns each run into a writing lesson, so your drafts need less fixing over time.

WriteOpen prompt

Generate headline variants that make one promise

Four fixed modes force real variety instead of twelve rewordings of one idea, which is what a bare 'give me headlines' prompt returns. The no-gap rule keeps variants aligned with what the piece can deliver, which protects credibility.

BrainstormOpen prompt

Run a structured editing pass

Editors work in ordered passes because sentence polish is wasted on paragraphs that will be cut; this prompt encodes that ordering. Reporting findings instead of rewriting keeps your voice intact and makes each fix a deliberate decision rather than a silent substitution.

AnalyzeOpen prompt

Structure a story before drafting it

The want-versus-need gap and the passive-beat check are the two diagnostics that catch most broken drafts before they are written. Keeping the output at structure level means you spend revision effort on architecture, where it is cheap, instead of on finished prose, where it is painful.

PlanOpen prompt

Draft a newsletter issue readers finish

Leading the subject line with a payload instead of a theme is the highest-leverage open-rate fix. The cut-if-no-why-care rule keeps issues short and reader-centric, and the forwarding-colleague voice instruction reliably strips newsletter-brand mush from the draft.

WriteOpen prompt

Draft a case study from messy notes

Every-number-needs-a-baseline is what separates evidence from marketing gas, and the [NEED] markers turn missing data into a follow-up list instead of silent invention. Including one real friction point is a credibility device readers consistently reward.

WriteOpen prompt

Shift the tone without losing the message

Extracting the load-bearing content first makes tone a layer instead of a rewrite, which is how professional editors handle register changes. Producing a restrained second variant compensates for the model's tendency to caricature the requested tone.

WriteOpen prompt

Expand an outline into a first draft

The per-section word budget and the no-preview rule prevent the two classic expansion failures: bloat and repeated framing. [EVIDENCE?] markers keep the model from padding your outline's gaps with plausible inventions, which is where AI drafts lose trust.

WriteOpen prompt

Write the title tag and meta description

Titles get truncated around 60 characters and descriptions around 160, so the caps are physical, not stylistic. Query-near-the-front matches how results pages bold terms and how people scan them; five deliberately different pairs give you a real choice instead of one guess.

WriteOpen prompt

Check a draft against your style guide

Quoting rule and offense together makes each finding verifiable in seconds, which is what makes automated review trustworthy. The guide-gap section turns the check into a feedback loop for the style guide itself instead of pure one-way enforcement.

AnalyzeOpen prompt

Turn scattered notes into an article

The inventory step is what stops note-to-article prompts from producing an essay loosely inspired by your notes: everything in the draft traces to something you wrote. Choosing the thesis the notes can support, rather than the one you hoped for, is the editorial judgment that makes the piece hold.

WriteOpen prompt

Marketing

12 prompts

Write an SEO content brief

Intent-first briefs prevent the classic failure of writing a great page for the wrong query type. Pre-banning the likely padding and demanding a named differentiator forces the brief to make editorial decisions, which is the entire value of a brief over a keyword list.

PlanOpen prompt

Write a first-touch cold email

The could-not-be-sent-to-anyone-else test on the first line is the single strongest filter against template smell. Hypothesis framing for the problem keeps the email respectful when your guess is wrong, and the micro-ask converts because it costs the reader almost nothing.

WriteOpen prompt

Draft landing page copy

The order mirrors how visitors actually read: outcome, proof, recognition, mechanism, objections. Answering the three real reasons for not buying beats another feature list, and the [VERIFY] convention keeps aspirational copy from shipping as fact.

WriteOpen prompt

Generate ad headline variants

The 30-character cap with visible counts matches real search-ad limits, so nothing gets written twice. Spreading variants across five named mechanisms produces a testable matrix instead of fifteen near-duplicates, which is what makes the A/B results interpretable.

BrainstormOpen prompt

Repurpose one piece into platform-native posts

Repurposing fails when one caption gets pasted everywhere; forcing a platform-native form and a restated lead per platform is the fix. Requiring one concrete detail per post keeps the variants anchored to the source instead of dissolving into generic advice.

WriteOpen prompt

Draft a press release journalists can use

Journalists cut from the bottom, so inverted-pyramid ordering decides what survives. Banning excited-to quotes and requiring an opinion with stakes produces the one part of a release that ever gets quoted verbatim.

WriteOpen prompt

Plan a product launch week by week

Working backward from a fixed date converts wishes into deadlines, and the single-claim consistency check catches the most common launch failure: five channels telling five stories. The day-after metrics with thresholds make the retro objective before the launch happens.

PlanOpen prompt

Draft evidence-based customer personas

Personas rot when they are demographic fiction; anchoring every attribute to cited evidence and quarantining assumptions keeps them empirical. The decision-it-changes test is the filter that separates working personas from slide decoration.

AnalyzeOpen prompt

Write an email nurture sequence

Assigning each email one named job prevents the sequence from becoming five identical pitches. Value-first ordering earns the attention the final ask spends, and the real-deadline-or-none rule keeps the sequence out of the trust-burning fake-urgency pattern.

WriteOpen prompt

Map competitor positioning to find open ground

Extracting the enemy each competitor picks reveals strategy faster than feature comparison. Making the model justify the axes prevents the lazy quality-vs-price map, and demanding evidence for claimable positions keeps the recommendation from being an empty white-space fantasy.

AnalyzeOpen prompt

Summarize a campaign debrief

Separating targeted metrics from observed ones blocks retroactive goal-setting, the most common debrief distortion. The bad-bet versus bad-execution split routes each failure to the right fix, and will/stop phrasing turns lessons into commitments instead of sentiment.

SummarizeOpen prompt

Cluster keywords into a page plan

Clustering by shared answer rather than shared words is the actual skill of information architecture, and the one-page-per-intent rule prevents the self-cannibalization that quietly caps rankings. Acceptance tests per page give writers a definition of done that is about the searcher.

PlanOpen prompt

Engineering

12 prompts

Review a diff like a senior engineer

The pageable-owner frame shifts review from style commentary to failure hunting, and the fixed severity ladder keeps nits from drowning blockers. Ordering correctness before style mirrors how strong reviewers actually allocate attention.

AnalyzeOpen prompt

Turn a vague bug into a reproducible report

Confining hypotheses to their own section keeps the factual report trustworthy while still capturing debugging intuition. The [GAP] markers turn missing reproduction steps into a checklist instead of fabricated certainty, which is what makes another engineer able to act on it.

WriteOpen prompt

Write SQL from a plain-language request

Most wrong SQL is a correct answer to a misread question, so the grain-and-edges restatement catches errors before they cost a table scan. The never-invent-columns rule plus the assumptions list makes silent schema hallucination visible.

CodeOpen prompt

Build a regex with tests, not vibes

Regexes fail at their edges, so the adversarial test table is the real deliverable; the regex is almost a byproduct. The not-regular escape hatch stops the classic disaster of forcing grammar-shaped problems into a pattern that almost works.

CodeOpen prompt

Plan a refactor you can ship in slices

Every-slice-ships-alone is the discipline that separates refactoring from a rewrite wearing a refactor's clothes. Naming the not-fixing list and the stop-the-line signals contains the two classic failure modes: silent scope creep and sunk-cost continuation.

PlanOpen prompt

Review an API design before it ships

Writing actual client code against the design is the strongest review technique in the list: every guess the reviewer makes is a place real integrators will guess differently. The consumer-who-cannot-change-it frame weights irreversibility correctly.

AnalyzeOpen prompt

Write a commit message from a diff

Behavior-over-mechanics subjects are what make git log a usable history, and the why-plus-rejected-alternatives body is the part future maintainers actually search for. The split proposal gently enforces atomic commits where the diff betrays a mixed change.

WriteOpen prompt

Derive test cases from a spec

Deriving the matrix before writing code keeps coverage systematic instead of vibes-driven, and the [SPEC?] discipline routes ambiguity back to the spec author instead of baking the model's guess into your test suite as truth. Hostile-QA cases surface the gaps specs systematically miss.

CodeOpen prompt

Write an architecture decision record

ADRs exist for the future engineer asking why; writing rejected options at full advocacy strength is what makes the record credible to them. The reopen trigger converts the decision from dogma into a falsifiable bet, which is how good teams treat architecture.

WriteOpen prompt

Draft a blameless incident postmortem

System-allowed framing is what keeps postmortems generative instead of defensive, and typed action items with falsifiable done-conditions are the difference between learning and theater. The uncomfortable-question section surfaces the structural issue most postmortems politely bury.

WriteOpen prompt

Outline a zero-drama migration plan

Expand/contract phasing with explicit entry/exit criteria is the pattern that removes the big-bang moment where migrations die. Requiring a measured agreement threshold before cutover replaces confidence with evidence, and the rollback rule bans the steps that cause resume-generating events.

PlanOpen prompt

Explain legacy code before touching it

Chesterton's fence, operationalized: the archaeology section explains the weirdness before you delete it, and characterization tests pin actual behavior (bugs included) so your change diffs cleanly against reality. The trap list focuses attention where your specific change meets their specific landmines.

AnalyzeOpen prompt

Business

12 prompts

Summarize a meeting transcript

Splitting decisions from open questions is the summary's real job; most meeting notes blur them, and work stalls in the blur. The no-interpolation and [UNCLEAR] rules keep the summary auditable against the transcript, which is what makes people trust it enough to skip meetings.

SummarizeOpen prompt

Extract action items from a transcript

The source-quote column makes every row verifiable in seconds, which is what lets a team adopt AI-extracted actions without re-listening. Separating firm from soft commitments preserves the real social texture of meetings instead of flattening maybes into assignments people never accepted.

ExtractOpen prompt

Write a weekly status report from raw updates

Outcome-verb discipline converts a diary into a report, and the blocker-must-carry-an-ask rule turns status into an unblocking machine. Committing next week in checkable phrasing builds the accountability loop that makes reports worth writing at all.

WriteOpen prompt

Compress a strategy into one page

Strategy is choice under constraint, so the will-not-do section is mandatory: without it a memo is a wish list. Demanding a differentiated belief and dated falsification signals forces the document to be a bet someone can hold you to, which is the entire function of writing strategy down.

WriteOpen prompt

Run a SWOT that produces decisions

A bare SWOT grid is a parking lot; the collision matrix (TOWS, formally) is what converts inventory into moves. Requiring falsifiable phrasing and evidence tags per item blocks the motivational-poster entries that make most SWOTs useless.

AnalyzeOpen prompt

Draft OKRs that measure outcomes

Converting outputs to the outcome each output should move is the core OKR discipline, and the self-audit for KR-hits-while-objective-fails catches the misalignment that makes teams cynical about the whole ritual. Anti-goals with floors prevent the classic quarter where growth eats quality.

PlanOpen prompt

Draft a board update that respects the room

Lowlights-first ordering is deliberate: boards calibrate trust on how you handle bad news, and burying it reads as either blindness or spin. Off-plan numbers paired with why-plus-response convert a metrics dump into evidence of a team in control.

WriteOpen prompt

Outline a proposal around the client's decision

Selling the after-state before the method matches how buyers actually decide, and phased deliverable framing turns your effort into their milestones. The exceeds-what-you-can-promise flags act as a contract-risk review before the proposal ever hardens into commitments.

PlanOpen prompt

Summarize a contract in plain language

Clause citations keep the summary tethered to the document instead of to contract norms the model has seen elsewhere, and the asymmetry hunt surfaces the terms most worth negotiating. Framing the output as lawyer-preparation sets the safe scope for AI on legal text.

SummarizeOpen prompt

Build a decision matrix that resists gaming

Constraint/criteria separation and the weight-sensitivity flip test are what keep a matrix from being decorated intuition. The disappointment check operationalizes a real decision-science trick: your reaction to the result reveals preferences the criteria list missed.

AnalyzeOpen prompt

Prepare for a meeting in five minutes

Per-attendee want/fear mapping converts an agenda into a negotiation plan, and pre-drafted answers to hard questions are what keep you composed when they arrive. The named-conflict rule forces preparation to face the real disagreement instead of hoping it stays quiet.

PlanOpen prompt

Summarize a quarter for the review meeting

The silently-abandoned category is the load-bearing part: most quarterly slippage is never decided, just unattended, and naming it changes planning behavior more than any velocity chart. Ending with decisions for the meeting makes the review forward-facing instead of a eulogy.

SummarizeOpen prompt

Education

12 prompts

Build a lesson plan around one misconception

Designing the lesson as a collision with a named misconception is how learning science says concepts actually change; content-coverage plans leave the wrong model intact underneath. Observable objectives and per-segment checks make 'did it work' answerable during the lesson, not at the exam.

PlanOpen prompt

Generate a rubric students can self-check against

Adverb-free level descriptions are the difference between a rubric and a vibe: levels described by observable behavior grade consistently across raters and days. The self-check rewrite turns the rubric from a grading tool into a teaching tool at zero extra cost.

WriteOpen prompt

Generate a quiz from course material

Distractors built from named misconceptions turn wrong answers into diagnostic data about who believes what. Requiring novel examples for application questions closes the loophole where students pass by pattern-matching the textbook's exact sentences.

WriteOpen prompt

Explain a concept to a complete beginner

Stating where the analogy breaks is the single highest-value instruction: analogies teach fast and mislead faster, and flagging the break preserves the speed without the damage. The dont-skip-the-hard-part rule counters the model's tendency to smooth difficulty away instead of decomposing it.

WriteOpen prompt

Compress notes into a study guide

Retrieval-format questions exploit the testing effect, the most replicated result in learning science; recognition-style summaries feel productive and test poorly. The [NOT FROM YOUR NOTES] marking keeps the guide faithful to your course's actual framing, which is what the exam grades.

SummarizeOpen prompt

Give essay feedback a student can act on

One-big-thing feedback respects the working-memory limits that make ten-item feedback useless, and demonstrate-once-then-point transfers the skill instead of the sentences. Deferring surface errors mirrors how writing instructors sequence revision: structure first, polish last.

AnalyzeOpen prompt

Outline a course from outcomes backward

Designing the capstone before the modules (backward design) guarantees the course teaches toward proof rather than coverage. The build-something-early and drop-curve requirements attack the real failure mode of courses, which is abandonment, not confusion.

PlanOpen prompt

Write discussion questions that start arguments

The two-defensible-answers test is what separates discussion from recitation, and supplying the concrete case inside application questions keeps the debate anchored instead of hypothetical. Facilitator follow-ups make the set usable by someone leading a discussion for the first time.

BrainstormOpen prompt

Draft a parent update that gets read

One specific dinner-table detail beats three paragraphs of curriculum summary for making families feel connected, and the single sub-10-minute action gets done where a list of five gets skipped. Plain statement of problems builds the trust that makes the hard conversations easier later.

WriteOpen prompt

Differentiate one activity three ways

Same-destination-different-ramp is the differentiation principle that keeps expectations high for everyone; three separate activities quietly become three tracks. The dignity check guards the classic failure where scaffolding removes the cognition and the supported tier becomes copying.

BrainstormOpen prompt

Outline a syllabus students actually use

The thread column and the student-facing promise reframe the syllabus from contract to map, which is how the students who succeed actually use it. Pre-answering the six inevitable emails and reasoning the AI policy line reduce friction all semester for one hour of design.

PlanOpen prompt

Generate analogies and stress-test them

Naming the load-bearing structure first forces analogies that preserve the right relation instead of surface resemblance. Pricing each analogy by the wrong conclusion it invites, then installing the boundary in the same breath as the analogy, is how good teachers use them without collateral damage.

BrainstormOpen prompt

Data

12 prompts

Plan a data cleaning pass before touching the data

Profile-before-fix stops the classic error of cleaning problems the data does not have while missing the ones it does. Reversibility labels and the quarantine rule preserve the audit trail that separates defensible analysis from data laundering.

PlanOpen prompt

Plan an exploratory analysis that answers a question

Pre-registering what each pattern would mean is the discipline that prevents just-so stories after the charts render. The confound list and sanity gates front-load the skepticism that usually arrives too late, in the meeting where someone asks about selection bias.

PlanOpen prompt

Choose the right chart for the message

Grounding chart choice in perceptual ranking (position > length > angle > area) replaces taste debates with mechanism. Titling the chart with the message and running the lie check are the two moves that most raise chart quality in real decks, and the final rule keeps the chart faithful to the data.

AnalyzeOpen prompt

Explain what a SQL query really does

Execution-order walkthroughs match how SQL actually behaves rather than how it reads, which is where most misreadings start. The silent-behaviors section names the three mechanisms (join fan-out, NULL logic, quiet row loss) behind nearly every dashboard number that turned out wrong.

AnalyzeOpen prompt

Document a dataset so others can trust it

Forcing a null-meaning and a gotcha entry per column is what turns a schema dump into documentation; those two columns hold everything analysts learn the hard way. The fitness statement sets the boundaries that prevent the dataset's numbers from wandering into claims they cannot support.

WriteOpen prompt

Define a metric before it defines you

Metrics rot at their edges, so the explicit edge rulings with rationale are the core of the document, not the formula. Pre-computing the gaming vectors and pairing each with a guardrail metric is what makes the metric safe to put in an incentive.

WriteOpen prompt

Read out an A/B test without fooling yourself

Verdict-first with the deciding number keeps readouts from burying the decision in statistics theater. The peeking flag, post-hoc segment quarantine, and believability check encode the three most common ways teams convince themselves of effects that are not there.

AnalyzeOpen prompt

Investigate a metric anomaly systematically

Artifact-first ordering reflects the base rates: most metric anomalies are measurement, not reality, and theorizing before ruling that out wastes days. The co-movement test is the cheapest hypothesis killer in analytics and almost nobody runs it systematically.

PlanOpen prompt

Specify a dashboard around decisions

Decision-first specification inverts the usual chart-collection process and is why some dashboards run companies while most decorate them. The kill list and the 30-day deletion test build in the lifecycle discipline that keeps dashboard sprawl from reaccumulating.

PlanOpen prompt

Summarize survey results without flattering them

Leading with response reality and demanding a base for every percentage inoculates the readout against the overclaiming that makes survey data untrusted. Typifying verbatims and minimum cell sizes keep the qualitative and subgroup layers evidence-grade.

SummarizeOpen prompt

Expose the assumptions inside a forecast

Sensitivity-ranked assumptions with plausible-pessimistic scenarios produce planning conversations instead of the dismissal that doom scenarios trigger. The correlation-trap section fixes the standard modeling sin of stress-testing inputs one at a time when reality moves them together.

AnalyzeOpen prompt

Audit data quality before anyone quotes it

Layered checks catch different failure classes: structural breaks are loud, semantic contradictions are quiet, and reconciliation against independent systems is the only check that catches a pipeline lying consistently. Thresholds and block-decisions convert quality from a feeling into an operating rule.

AnalyzeOpen prompt

Sales

12 prompts

Prepare a discovery call that earns the second call

Framing the problem as a hypothesis to test keeps discovery from collapsing into a pitch, and the SPIN-ordered ladder builds the implication trail that makes budget conversations natural later. Explicit disqualification criteria protect the pipeline from optimistic clutter.

PlanOpen prompt

Write the follow-up email after a demo

Opening with their words and mapping bullets to voiced priorities proves listening, which is the actual differentiator in follow-ups. The single dated next step converts momentum into calendar reality, and the skeptic P.S. works the deal's real decision dynamics instead of just its champion.

WriteOpen prompt

Draft responses to a sales objection

Diagnose-then-respond beats scripted rebuttals because most objections are proxies; the clarifying question routes you to the real one. Mandatory concession of the true part is the credibility move that separates a consultative answer from a script, and buyers can tell in one sentence.

WriteOpen prompt

Write the executive summary that sells the proposal

Economic buyers read one page and skim the rest, so the summary carries the deal: situation-cost before solution mirrors how they justify spend internally. Placing investment beside status-quo cost reframes price as a comparison you chose, and the their-name-count rule enforces buyer-centricity mechanically.

WriteOpen prompt

Build an account research brief

Trigger-mapped research answers the only question that matters at the top of an account: why now. Predicting the disrupted stakeholder alongside the buyer prepares you for the blocker most reps meet by surprise in month two of the deal.

AnalyzeOpen prompt

Write cold call openers that survive five seconds

The first five seconds decide whether the brain files you as intrusion or as relevance; trigger and permission openers survive because they concede what is true (you are interrupting) and prove specificity fast. Pre-scripting the next line for the three likely responses is what makes the plan survive contact.

BrainstormOpen prompt

Answer the pricing question without dodging

Buyers ask early to self-qualify, and a number with anchors respects that while keeping sizing control. The three-reaction follow-ups turn the scariest moment of early conversations into a planned fork, including the underpricing tell most reps miss.

WriteOpen prompt

Write a renewal check-in that is not an invoice warning

Renewals are decided by accumulated value evidence, not by the renewal email; leading with their numbers makes the case theirs. Naming weak adoption at day 90 converts the most dangerous renewal pattern (silent drift, then shock) into a recoverable working conversation.

WriteOpen prompt

Run win/loss interviews that tell the truth

Chronological walk-throughs recover the real decision sequence that summary questions flatten into rationalization, and the counterfactual question is where losses give up their lesson. The disconfirmation self-test keeps the guide from becoming a machine for confirming what sales already believes.

PlanOpen prompt

Outline a territory plan that survives the quarter

Territory plans fail as lists of accounts with vibes; the opening math forces the plan to be arithmetic before it is strategy. Tier-differentiated plays allocate scarce depth work where fit evidence justifies it, and the re-plan trigger prevents the silent months where a bad plan runs to zero.

PlanOpen prompt

Write LinkedIn outreach that does not smell like a sequence

The 200-people test is the sharpest filter for outreach that reads as human, and one-line-from-a-phone questions match how LinkedIn actually gets answered. Requiring new value per follow-up replaces persistence theater with earned attention, and the graceful exit protects a small-world reputation.

WriteOpen prompt

Write sales-to-success handoff notes

The volunteered-versus-suggested split on success criteria identifies the promises the renewal actually rides on, and the promise ledger with contractual/verbal/implied labels is where churn-causing surprises get caught early. The customer-could-read-it rule keeps candor from sliding into cynicism.

WriteOpen prompt

Design

12 prompts

Draft a design brief that prevents rework

Separating problem from smuggled solution preserves the design exploration a brief exists to enable, and pre-agreed success behavior prevents the taste-war reviews that cause rework. Naming the direction-vs-detail purpose of each checkpoint stops the classic failure of relitigating strategy at the polish stage.

WriteOpen prompt

Write UX copy variants for one moment

Posture-labeled variants make the review conversation about the user's emotional state instead of word preference, which is how UX writing decisions should be made. The per-element structure with character counts and verb-first buttons produces copy that drops into the design tool without a translation step.

BrainstormOpen prompt

Outline a user flow with its failure paths

Demanding failure-path parity with the happy path is where this prompt earns its keep: most flow documents die at the first error state, and users do not. The state inventory for mid-flow returns covers the resume experience that separates polished products from prototypes.

PlanOpen prompt

Write a usability test script that does not lead

Scenario-goal phrasing and the label-matching ban are what make a test measure findability instead of reading ability. Scripted neutral probes protect the moment moderators most often contaminate: the silence after a participant gets stuck.

WriteOpen prompt

Refine an idea into an image-generation prompt

Image models reward specificity in exactly the categories this prompt forces (subject, composition, light, palette, style), and lead-with-subject matches how prompt tokens are weighted. Planning the defect check and the one-parameter iteration turns generation from slot-machine pulls into a controlled search.

WriteOpen prompt

Find the words for a visual direction

Direction words with anti-words give a moodboard its rejection criteria, which is what makes curation converge instead of sprawl. Hunting references outside the product category and describing qualities instead of naming artists produces directions a team can own rather than imitate.

BrainstormOpen prompt

Review a screen for accessibility

Per-element verdicts with computed contrast ratios produce a fixable punch list instead of an awareness lecture. The cannot-assess candor matters: static reviews systematically miss focus order and announcement bugs, and saying which tool catches each keeps the checklist from breeding false confidence.

AnalyzeOpen prompt

Ask critique questions that improve the work

Stage-gating the feedback type prevents the two classic critique failures: pixel-polishing a concept and re-opening strategy during polish. The strong-answer-versus-rationalization guide turns questions into calibration for the whole room, and the closing ritual converts talk into decisions.

BrainstormOpen prompt

Outline a portfolio case study that shows judgment

Hiring reviews scan for judgment, and judgment lives in the fork section: real alternatives with reasons for the losers. Captioning artifacts with decisions instead of deliverable names reframes the whole study from output gallery to thinking record, which is the actual hiring signal.

PlanOpen prompt

Brainstorm icon concepts beyond the obvious

Naming the clichés first makes the use-or-avoid decision deliberate instead of accidental. The four-approach split, misread-risk-at-16px, and silhouette-based ranking evaluate concepts the way icons actually fail: small, fast, monochrome, and next to their siblings.

BrainstormOpen prompt

Define brand voice attributes with edges

Attributes with but-not edges and paired over-the-edge examples are what make voice guidance operational; bare adjective lists produce whatever the writer already writes. Testing the voice in hard moments (errors, incidents, price increases) defines it where brand trust is actually won or lost.

AnalyzeOpen prompt

Write design handoff annotations developers can build from

The states table and content min/max rules cover the two questions developers actually ping designers about, and [DECISION NEEDED] converts silent gaps into visible work items before sprint pressure fills them arbitrarily. Mapping to named system components makes cost visible where a new component is being smuggled in.

WriteOpen prompt

Personal productivity

12 prompts

Turn a task pile into a daily plan

The 50 percent estimate padding corrects planning fallacy with arithmetic instead of willpower, and the explicit full-line converts overcommitment from a private failure into a visible scheduling fact. The one-thing anchor gives the day a definition of success that survives interruptions.

PlanOpen prompt

Design inbox triage rules you will actually follow

If-then rules with time bounds remove the per-message decisions that make inboxes exhausting; the executable-while-tired test is the real filter. Separating triage from working stops the reply-spiral that turns 20 minutes of sorting into 3 hours of unplanned work.

PlanOpen prompt

Run a weekly review that changes next week

Pattern questions ('what should I have declined') produce behavior change where inventory questions produce lists, and the zombie-task kill-or-commit rule is the only reliable cure for task-list rot. Pre-decided trade-offs make next week's collisions cheap because the decision already happened.

PlanOpen prompt

Break a big goal into a workable plan

The math check refuses the polite fiction most plans are built on, forcing the real trade-off conversation on day one instead of at the deadline. Embarrassingly small first actions exploit how starting works, and pre-committed responses to the predicted week-3 dip make persistence a design property instead of a mood.

PlanOpen prompt

Build a learning plan for a new skill

Deconstruction plus feedback-source-per-sub-skill applies the two strongest findings in skill acquisition: decompose and shorten the error-correction loop. Scheduling a real output in week one and pre-banning the classic time sinks protects the scarce resource, which is motivation, not information.

PlanOpen prompt

Draft a travel itinerary with slack built in

One-anchor-plus-neighborhood-clustering removes the transit thrash that eats trip hours, and visible slack with named fallbacks keeps free time restful instead of decision-heavy. The booking checklist ordered by sell-out speed sequences the only part where timing is unforgiving.

PlanOpen prompt

Plan a week of meals around real life

Effort-matched-to-calendar is what makes a meal plan survive contact with an actual week, and the ingredient bridge cuts both cost and waste without recipe gymnastics. The escape hatch and the specialty-ingredient rule handle the two ways plans die: collapse nights and one-use jars.

PlanOpen prompt

Prioritize a reading list by decision value

Sorting by connection to live decisions converts a guilt pile into an instrument, and the read-to-answer-a-question technique is the highest-leverage retention move available. The summary-instead pile prices books at their idea density, and the intake rule fixes the upstream habit that built the pile.

AnalyzeOpen prompt

Design a habit system, not just a tracker

Minimum-viable reps with streaks counting the minimum decouple identity from intensity, which is what makes habits survive bad weeks. Trigger-attachment beats time-scheduling in the habit literature, and the never-two protocol treats the real killer (the post-miss spiral) as a design problem rather than a character problem.

PlanOpen prompt

Generate journaling prompts that go somewhere

Mode variety is the engine: specific-recall grounds, pattern-hunt extracts, reframe unsticks, forward converts insight into motion, and rotating them prevents the rumination loop that gives journaling its mixed reputation. Follow-ups for generic first answers push past the diary's small talk.

BrainstormOpen prompt

Create a decision journal template

Recording probability, mood, and rejected alternatives at decision time is what makes later review reliable; memory rewrites all three within weeks. The process-versus-luck question at review is the core of decision-quality practice, separating what you control from what happened to you.

WriteOpen prompt

Plan a deep focus session

The on-ramp task defeats the start-resistance that costs the first 20 minutes, and the capture pad converts distraction from a derailment into a one-line note. Landing notes exploit the cheapest productivity trick that exists: tomorrow's momentum is manufactured during the last five minutes of the current session.

PlanOpen prompt

More free resources

Free toolsCalculators and checkers on live model dataTemplatesAgent, automation, board, and skill setupsLearnPlain-language definitions and benchmark explainers

Ready to get to work?

Free to start. No credit card. Never trained on your data.

Start free
  • Accueil
  • Tarifs
  • Modèles IA
  • Modèles d'image
  • Modèles de voix
  • Modèles vidéo
  • Classements
  • New models
  • Model status
  • Multi-Model Chat
  • Voice Mode
  • Voice HUD
  • Web Search
  • Image Generation
  • Video Generation
  • Audio Generation
  • Transcription
  • Drive
  • Secrets
  • Sharing
  • Workspaces
  • Tasks
  • Memory
  • Agents
  • Subagents
  • Automations
  • Skills
  • Code Execution
  • Computers
  • Computer Use
  • Cloud Computers
  • Local AI
  • AI Gateway
  • API & SDK
  • CLI
  • MCP
  • Tunnels
  • All features →
  • Blog
  • Cas d'usage
  • Comparisons
  • Best of
  • Learn
  • Free tools
  • Prompts
  • Templates
  • Changelog
  • Centre d'aide
  • FAQ
  • Confidentialité
  • Comparer tous les modèles
  • Support
  • Développeurs
  • Quickstarts
  • Référence API
  • Tarifs API
  • CLI
  • MCP
  • Downloads
  • Desktop
  • Badges and embeds
© idapt[email protected]ConditionsPolitique de confidentialitéMentions légalesSignaler un contenu
X (Twitter)