Best Prompts for your tasks.

You are a senior [language] engineer reviewing a pull request on a [system type]. Our team values [priorities]. Here is the diff: <diff> [diff] </diff> Review it in this order and stop at what actually matters: 1. Correctness bugs, including edge cases, null/empty inputs, and concurrency issues. 2. Security and data-safety risks. 3. Performance problems that would show up at [expected scale]. 4. Readability and convention drift. Output a markdown table with columns: Severity (blocker / should-fix / nit), File:line, What is wrong, Concrete fix. Then write a two-sentence verdict on whether this should merge. Do not comment on style the linter already enforces, and do not invent problems if the diff is clean.

Act as the most thorough reviewer on my team. I am about to submit this [language] change that [what it does]. Context you need: [context]. <code> [code] </code> Before listing anything, restate in one sentence what you believe this code is supposed to do. If that differs from my description, say so first — that mismatch is the most important finding. Then give me: (a) the three things a reviewer is most likely to reject this for, (b) any test case I am missing, and (c) one simplification that would remove code without changing behavior. For each point, quote the exact lines and show the replacement. Be direct; skip praise.

You are an application security engineer. Review the following [language] code, which handles [data or action] in a [app type] running on [environment]. <code> [code] </code> Walk the untrusted input from where it enters the system to where it is used. For each place it is trusted without validation, report: the input, the sink it reaches, the realistic attack, and the minimal fix in code. Check specifically for injection, broken access control, unsafe deserialization, secret leakage into logs, and missing rate limits. Format as a numbered list ordered by exploitability. End with 'No issues found in scope' if that is the honest answer, and say what you could not evaluate without seeing more of the codebase.

You are a test engineer writing [test framework] tests for a [language] codebase. Match this style: [style example]. Function under test: <code> [function] </code> First, list the behaviors worth testing as short bullet points: happy path, boundary values, invalid input, error paths, and any state or side effects. Ask yourself which of these a bug would realistically hide in, and mark those as priority. Then write the tests, one assertion concept per test, with names that read as sentences describing the behavior. Use fixtures or mocks only where a real dependency would make the test slow or flaky. Return only the test file, ready to run, followed by a one-line note on any behavior you could not test and why.

Here is a [language] function and its existing tests. <function> [function] </function> <tests> [existing tests] </tests> Act as a QA engineer whose job is to break this code. Ignore line coverage; think about input space. Identify inputs and sequences that are currently untested and could produce wrong output, an unhandled exception, or corrupted state — empty and huge inputs, unicode, negative and zero values, duplicate calls, out-of-order calls, timezone and locale differences, and concurrent access if relevant. Output a table: Scenario, Why it could break, Expected behavior. Then write the three highest-value missing tests in full. If the existing tests already cover the risky cases, say so plainly instead of padding the list.

I have a bug: [symptom]. It appears in this code: <code> [code] </code> Environment: [environment]. Write a single minimal [test framework] test that fails today because of this bug and will pass once it is fixed. The test must assert the correct behavior, not the buggy behavior. Keep setup to the smallest thing that reproduces it. After the test, explain in three sentences what the test proves and what root cause you suspect. Do not fix the code yet — I want the failing test first so the fix is verifiable.

You are a staff engineer onboarding me to an unfamiliar codebase. I am fluent in [languages I know] but have never seen [framework or domain]. <code> [file or module] </code> Explain it in this order: 1. One paragraph: what business problem this code solves and who calls it. 2. A walkthrough of the main flow, step by step, in plain language, referencing function names. 3. The non-obvious parts — anything that looks wrong but probably exists for a reason, plus your best guess at that reason. 4. Inputs, outputs, side effects, and external systems touched. 5. Where I would most likely break something if I edited this. Mark anything you are inferring rather than reading directly as 'assumption'. Keep it under 600 words.

Act as a technical writer who also codes. Produce documentation for this [language] module so the next engineer does not have to read the source. <code> [code] </code> Write it as markdown with: a one-sentence purpose, a usage example that would actually run, a table of public functions (signature, what it does, gotchas), a short 'how it fits in' section describing upstream callers and downstream dependencies, and a 'known limitations' section based only on what the code shows. Use the vocabulary already present in the code rather than inventing new terms. Where the code's intent is genuinely ambiguous, add a 'TODO: confirm with the team' note instead of guessing confidently.

You are a senior engineer who has been paged at 3am for changes like this one. Here is a legacy [language] file we need to change to support [new requirement]. <code> [code] </code> Constraints: [constraints]. Give me: (1) the smallest change that satisfies the requirement without restructuring anything, (2) the characterization tests I should write first to lock in current behavior, (3) a step-by-step refactor sequence where each step is independently shippable and reversible, and (4) the parts of this file I should leave alone and why. Be explicit about which risks are real and which are theoretical. Prefer the boring option.

You are debugging with me, not for me. Here is what I have. Stack trace / error: <error> [error] </error> Relevant code: <code> [code] </code> Environment: [environment] What changed recently: [what changed] What I already tried: [what you tried] Give me the three most likely root causes ranked by probability, and for each: the evidence in the trace supporting it, one command or log line that would confirm or eliminate it in under five minutes, and the fix if confirmed. Do not suggest fixes that only suppress the error. If the trace is insufficient, tell me exactly what to capture next.

Act as a systems debugger. I have a bug that reproduces roughly [frequency] in [environment] but never locally. Symptom: [symptom]. Stack: [tech stack]. Relevant code and config: <context> [context] </context> Generate hypotheses grouped by category — race conditions and shared state, timeouts and retries, caching and staleness, resource exhaustion, environment/config differences, clock and timezone, and third-party flakiness. For each hypothesis, state the specific observable signature I should look for in logs or metrics. Then give me an ordered debugging plan where each step either narrows the space or is cheap to run. End with the single instrumentation change most likely to catch this in the act.

Explain this error to me as if I am competent but new to [technology]. <error> [error] </error> Context: I was trying to [what you were doing] in [project type]. Answer in four short parts: what the message literally means, what usually causes it in real projects, the two or three most common fixes with exact commands or code, and how to tell which one applies to my case. Avoid jargon unless you define it in the same sentence. If the error is a symptom of something upstream, say what that upstream thing usually is.

You are a senior marketing strategist. Write a campaign brief for [product or service] targeting [audience] with a budget of [budget] over [timeframe]. What you know: our positioning is [positioning], our main competitor is [competitor], the business goal is [goal and metric], and the constraint is [constraint]. Structure the brief as: Objective (one measurable sentence) · Audience insight (what they actually believe today and what we need them to believe) · Single-minded proposition · Reasons to believe (max three) · Channel plan with rough budget split and rationale · Key messages per channel · Success metrics with target numbers · Risks and what we will not do. Keep it to one page. Flag any place where you had to assume something I did not tell you.

Act as a skeptical CMO reviewing a campaign proposal. Here it is: <brief> [campaign idea] </brief> Budget: [budget]. Goal: [goal]. Audience: [audience]. Challenge it honestly: Is the audience insight real or a stereotype? Does the creative idea depend on the audience caring more than they do? Is the budget split defensible? What is the single assumption that, if wrong, sinks the whole thing? Give me: three hard questions I should be able to answer before spending money, the weakest part of the plan with a specific fix, and one cheaper test that would validate the core assumption in under two weeks. Be blunt — I would rather hear it now than after launch.

You are a marketing operations lead. Convert this campaign brief into an execution calendar. <brief> [brief] </brief> Launch date: [launch date]. Team: [team]. Channels: [channels]. Output a table with columns: Week, Deliverable, Owner role, Dependency, Status gate. Work backward from launch so that dependencies (creative approval, legal, asset production, media booking deadlines) land before the things that need them. Add a pre-launch checklist and a post-launch measurement schedule for days 1, 7, and 30. Call out any deliverable that is unrealistic for the stated team size, and propose what to cut rather than silently overloading the plan.

You are a social media manager for [brand], which sells [product] to [audience]. Our brand voice is [voice]. Active channels: [channels]. Posting cadence: [cadence]. Build a one-month calendar around the theme [theme or campaign]. Use a content mix of roughly 40% educational, 30% social proof or customer stories, 20% behind-the-scenes or personality, 10% direct promotion. Output a table: Date, Channel, Format (carousel/short video/text/photo), Hook line, Post copy (platform-appropriate length), CTA, Asset needed. Write real copy, not placeholders. Avoid hashtag stuffing and any phrase we would be embarrassed to read aloud. Finish with three post ideas you think are risky but could outperform.

Act as a content strategist. Here is a long-form asset we published: <asset> [asset] </asset> Brand voice: [voice]. Audience: [audience]. Channels: [channels]. Extract the five most quotable or counterintuitive ideas in it. Turn each into a native post for the channel where it fits best — not the same text reformatted, but rewritten for how people actually read on that platform. Include a scroll-stopping first line, the body, and a CTA that matches the awareness level of that audience. Output one block per post labeled with channel and format. Then list two ideas from the asset that are NOT worth posting and say why.

You are a social media analyst. Here are our recent posts with their metrics: <data> [posts] </data> Audience: [audience]. Goal: [goal]. Identify the pattern separating the top quartile from the bottom quartile — hook style, length, format, topic, posting time, or CTA. Say which of these the data actually supports and which would need more posts to conclude. Then rewrite the three worst performers using what the winners did, keeping the original intent. For each rewrite, name the specific change you made and what you would expect it to move. End with two concrete tests for next month, each with a hypothesis and the metric that decides it.

You are an email marketer writing for [brand], which helps [audience] to [outcome]. Voice: [voice]. The reader just [trigger]. Write a [number]-email sequence over [timeframe]. Each email should earn the next one: lead with something useful, not a pitch. Email 1 delivers value tied to the trigger; the middle emails handle the top objections ([objections]); the last makes one clear ask. For each email give: send day, subject line plus one alternative, preview text, body under 150 words, and a single CTA. Write at a grade 7 reading level, use short paragraphs, and no phrases like 'in today's fast-paced world'. End with a note on what to personalize per segment.

Act as a conversion copywriter. Here is my email: <email> [email] </email> Audience: [audience]. Their current problem: [problem]. Goal of this email: [goal]. Brand voice: [voice]. Write 10 subject lines under 50 characters, deliberately spread across different angles: curiosity, specific benefit, question, number/data, social proof, urgency (only if honest), and plain-and-direct. Pair each with preview text that extends rather than repeats it. Then pick the two you would actually A/B test, explain what each one tests about the audience, and name the metric that decides the winner. Flag any line that risks landing in spam filters or overpromising relative to what the email delivers.

You are a direct-response copywriter. This email gets [open rate] opens but only [click rate] clicks: <email> [email] </email> Audience: [audience]. Desired action: [action]. Offer: [offer]. Diagnose first: is the problem the offer, the relevance, the clarity of the ask, the friction after the click, or the mismatch between subject line and body? Say which, with the evidence from the copy itself. Then rewrite it. Keep the same offer, cut the word count by at least a third, put the reader's problem in the first line, and make the CTA a single unmistakable action. Show the rewrite, then a short list of what you changed and why.

You are a competitive strategist. Compare our positioning against these competitors. Us: [us]. Competitors: [competitors]. Build a table: Competitor, Who they target, Core promise, Proof they offer, Price posture, Where they are genuinely better, Where they are vulnerable. Then answer three questions directly: (1) What claim can we make that none of them can credibly make? (2) Which of our current messages are commodity claims everyone makes? (3) If I were their head of marketing, how would I attack us? Base everything on the information I provided; label any market context you add from general knowledge as such, and do not invent statistics.

Act as a messaging analyst. Here is a competitor's website copy, ads, or sales material: <content> [competitor copy] </content> Our product: [description]. Our audience: [audience]. Reverse-engineer their strategy: which customer are they actually speaking to, what problem do they lead with, what objection are they pre-handling, what proof do they lean on, and what are they conspicuously silent about? Quote the specific lines that reveal each. Then give me: three things they do better that we should honestly steal, two claims of theirs that are weak enough to contrast against, and one message we could own because their positioning prevents them from saying it. Keep it under 500 words.

You are a competitive intelligence analyst. Here are notes from recent won and lost deals: <notes> [notes] </notes> Cluster them into themes. For each theme report: how often it appears, whether it shows up in wins, losses, or both, the exact customer language used, and whether it is a product gap, a pricing issue, a messaging failure, or a sales-process problem. Rank the themes by how much revenue they appear to touch. Then recommend the one marketing change and the one sales-enablement asset most likely to move the biggest theme. Be explicit about which conclusions are well supported by this sample and which are based on only one or two data points.

You are an experienced [subject] teacher planning for [grade level]. Standard or objective: [standard]. Class length: [minutes]. Class context: [class context]. Write a lesson plan with: learning objective in student-friendly language · a 5-minute hook that connects to something students already care about · direct instruction with the two or three key ideas and how to explain each · a guided practice activity · independent practice · a 3-question exit ticket that reveals whether the objective was met · timings for each segment that add to the full period. Include one common misconception students have about this topic and how to address it when it appears. Keep materials to what a normal classroom actually has.

Act as an instructional coach. Here is my lesson plan for [grade] [subject]: <plan> [plan] </plan> Adapt it for three groups in the same room: students who are below grade level in [specific skill], students on level, and students who will finish early and need depth rather than more worksheets. For each group give: the same core objective restated at the right level, the scaffold or extension, the specific task, and how I will know in the moment whether it is working. Also suggest supports for [ELL or IEP need] that do not single students out. Keep the whole thing runnable by one teacher with no aide, and tell me what to drop if we run short on time.

You are a curriculum designer. Build a [number]-week unit on [topic] for [grade] [subject], aligned to [standards]. Students arrive knowing [prior knowledge] and struggle with [known weakness]. We meet [frequency]. Sequence the unit so each week builds on the last. For each week give: essential question, the key concepts, one anchor activity, and the formative check. Then specify the summative assessment and a simple rubric. Build in two catch-up days and say what they are for. Flag the point in the unit where students most commonly get lost and what the recovery plan is. Present it as a table plus a short paragraph explaining the reasoning behind the sequence.

You are an assessment writer for [grade] [subject]. Create a [number]-question quiz on [topic], covering [learning objectives]. Mix the questions across Bloom's levels: about half recall and understanding, a third application and analysis, and a few evaluation or creation. Include [formats]. For every multiple-choice question, make the distractors reflect actual student misconceptions rather than obviously wrong answers, and note in the answer key which misconception each distractor targets. Provide the full answer key with a one-line explanation for each correct answer. Keep the reading level appropriate for [grade] and avoid cultural references that only some students would know. Mark the two questions most likely to be too hard.

Act as a teacher building an assessment strictly from the material students were given. Here it is: <material> [material] </material> Grade: [grade]. Time available: [minutes]. Purpose: [purpose]. Write [number] questions that can be answered from this material alone — no outside knowledge required. Vary the format and cover the whole text rather than the first page. Include at least two questions that require students to connect two different parts of the material. Output the student-facing questions first, cleanly formatted for printing, then a separate answer key with the exact line or paragraph from the source that supports each answer. Tell me if the material is too thin to support this many good questions.

You are an assessment designer. Create a rubric for this assignment: [assignment] for [grade] [subject]. It should measure [criteria]. Use four performance levels. Write the descriptors in second person and in language a [grade] student can self-assess with — concrete and observable, never vague words like 'good' or 'adequate'. Each level should describe what is present, not just what is missing. Output as a table, then add: a student-friendly checklist version they can use before submitting, and three examples of feedback comments I could give a student sitting at the second-highest level to move them up. Keep the whole rubric to one page.

You are an experienced [grade level] [subject] teacher grading student work. Use only the rubric below — do not apply your own standards. <rubric> [rubric] </rubric> <essay> [student work] </essay> For each rubric criterion give: the level earned, one sentence justifying it, and a direct quote from the essay as evidence. Then give an overall score and a comment written to the student in a warm, specific voice: two things they did well (with quotes) and the single highest-leverage thing to fix next time, with an example of how to fix it. Do not rewrite the essay for them. Do not comment on handwriting, effort, or anything the rubric does not measure. Flag anything that reads as possibly not the student's own work, without accusing.

Act as a writing coach giving feedback to a [grade level] student on a [assignment type]. The learning goal is [goal]. <essay> [student work] </essay> Give feedback in this order: what is working, one structural issue, one paragraph-level issue, and one sentence-level pattern (choose the most frequent error, not every error). Quote the student's own words when pointing to something, and phrase every suggestion as a question or an invitation to try something, not a correction. End with one revision task the student can complete in 15 minutes that would most improve the piece. Keep the whole response under 300 words and at a reading level the student can absorb without an adult translating it.

You are helping me grade a set of student responses consistently. Rubric: [rubric]. Prompt students answered: [prompt]. Here are the responses, separated by ---: <responses> [responses] </responses> For each response output a row: Student label, Score per criterion, Total, One-sentence comment. Apply the rubric identically across all of them; before scoring, state the specific bar you are using for the top level so I can check your calibration. After the table, summarize class-wide patterns: what most students got right, the two most common errors, and what I should reteach tomorrow. Note any response you found hard to score and why, so I can review it myself.

You are a [grade level] teacher writing to a parent or guardian about a concern. Facts: [facts]. Student strengths: [strengths]. Goal of the email: [goal]. Write an email that opens with something genuine and specific about the student, states the concern factually without labeling the child, describes what you have already done, makes one clear request, and offers two concrete times to talk. Keep it under 200 words, at a reading level accessible to any family, warm but not apologetic. Avoid educational jargon and any wording that could read as blame. Then give me a two-sentence version for a text message and a note on anything I should document.

Act as a veteran teacher preparing me for a conference with the family of [student]. Data I have: [data]. The family's likely concern: [concern]. My goal for the meeting: [goal]. Give me: a one-page talking-points sheet ordered as strengths, growth areas with evidence, what I am doing about it, and what would help from home · the three questions the family is most likely to ask and a direct answer for each · one question I should ask them · and how to redirect if the conversation turns adversarial. Use plain language a family with no education background can follow, and keep every claim tied to specific evidence rather than impressions.

You are a teacher who sends good news home deliberately. Write a short message to the family of a [grade level] student about [positive moment]. Make it specific enough that the family knows I actually noticed this child — name the behavior, the moment, and why it matters for their learning, not just that they were 'well behaved'. Keep it under 100 words, warm and direct, with no request attached. Give me three variations: one for email, one short enough for a text or app message, and one for a handwritten note. Then suggest a simple rotation so every family in a class of [class size] hears something positive at least once per [timeframe].

You are helping a nurse organize documentation. Do not provide diagnoses or treatment recommendations — structure only what I report. My raw notes: <notes> [rough notes] </notes> Organize them into a [note format] note using standard clinical abbreviations. Keep every clinical detail I gave you and add nothing. Where my notes are ambiguous or a required element appears missing (time, dose, route, patient response, follow-up), list it under 'Gaps to confirm before charting' rather than filling it in. Use objective, non-judgmental language and quote the patient directly where I recorded their words. Output the note, then the gaps list. Remind me that I am responsible for verifying accuracy against the record before entry.

Act as a documentation assistant for a nurse preparing an SBAR handoff. Use only the information I provide; do not infer clinical judgments I have not stated. <information> [patient info] </information> Write it as Situation, Background, Assessment, Recommendation — concise enough to deliver verbally in under 90 seconds. Put time-sensitive items first within each section. Then list, separately: anything the receiving nurse will predictably ask that my notes do not answer, and any element of a complete handoff that is missing. Keep language objective and free of speculation about causes.

You are reviewing nursing documentation for clarity and completeness — not for clinical decision-making. <note> [draft note] </note> Check it for: subjective or judgmental language that should be objective, vague quantifiers that should be measurements, missing timestamps, interventions recorded without the patient's response, and anything that reads as a conclusion rather than an observation. Return the revised note with changes marked, then a short list of what you changed and why. Do not add clinical content I did not write, and do not remove any detail — if something seems redundant, flag it instead of deleting. Finish by noting any element a chart auditor would likely question.

You are a nurse educator writing patient-facing material. Explain [topic] to a patient who is [patient context]. Write at a 6th-grade reading level. Use short sentences, everyday words, and 'you' throughout. Structure it as: what this is · why it matters for you · what you will do, step by step · what is normal to feel · when to call us right away (be specific about symptoms and numbers) · questions to ask your care team. Do not include dosing or individualized medical advice — leave [] placeholders for the clinician to fill in. Keep it to one page. End with a line reminding the patient to follow their own care team's instructions.

Act as a nurse writing discharge instructions. Clinical content to convey: [clinical instructions]. Patient context: [patient context]. Rewrite it so the patient can actually follow it at home. Organize by when they will need it: today, this week, ongoing. Turn every instruction into a concrete action with a time attached. Convert clinical terms into plain words, keeping the clinical term in parentheses once. Add a red-flag section written as 'Call [] right away if you notice...' with specific, observable signs. Use a checkbox format where a caregiver could tick items off. Do not add or change any clinical content — flag anything in the source that is unclear or contradictory instead.

You are a nurse using the teach-back method. The patient has just been taught about [topic]. Key points they must retain: [must-know points]. Write five teach-back questions that ask the patient to explain in their own words rather than answer yes or no, phrased so they do not feel tested — for example framing it as checking whether I explained it well. For each question, note what a complete answer contains and what a partial answer signals I should reteach. Then write a two-sentence re-explanation for the single point patients most often misunderstand about this topic. Keep all language at a 6th-grade level and culturally neutral.

You are helping a nurse organize a shift. Do not make clinical decisions — help me sequence work based on what I tell you. Assignment: <patients> [patients] </patients> Shift: [shift hours]. Fixed commitments: [fixed commitments]. Build an hour-by-hour plan that clusters tasks by room to reduce walking, respects fixed med times, and leaves buffer for the unpredictable. Mark which items are time-critical versus flexible, and which are safe to delegate to [role] if available. Then list the three things most likely to blow up this plan and what I should do first when they do. Note that clinical prioritization remains my judgment call.

Act as a documentation assistant. Convert my running notes from this shift into a clean report sheet for the oncoming nurse. <notes> [notes] </notes> For each patient produce: room and identifiers I provided, one-line status, key overnight events with times, current lines/drains/devices, pending items with who owns them, and specific things to watch for. Preserve every clinical detail exactly as I wrote it; do not smooth over uncertainty. Put anything urgent or time-bound at the top of each patient's block. At the end, list items that appear in my notes without a resolution so I can close them out before handoff. Do not add clinical interpretation.

You are helping write an objective, factual account of an event for internal reporting. Do not assign fault, speculate on cause, or add anything I did not state. What happened, in my words: <account> [account] </account> Rewrite it in chronological order using objective, observable language. Replace any conclusion with the observation behind it. Keep exact times, doses, and quotes. Use neutral phrasing for people involved. Output the account, then a list of factual gaps a reviewer would ask about (times, witnesses, notifications made, patient response). Remind me to follow my facility's reporting policy and to have this reviewed before submission.

You are assisting a nurse with the structure of a care plan. Use only the assessment data I provide; do not diagnose or prescribe. Assessment data: <data> [assessment data] </data> Setting: [setting]. Timeframe: [timeframe]. For each problem I have identified — [problems] — lay out the standard structure: supporting data from what I gave you, one measurable short-term goal with a timeframe, one long-term goal, candidate interventions with the rationale for each, and how the outcome would be evaluated. Write goals in patient-centered, measurable terms. Mark anything where my data is insufficient to support a goal. Present it as a table per problem, and note that problem selection and clinical judgment remain mine.

Act as a clinical educator reviewing care plan wording. Here are my goals and interventions: <plan> [goals and interventions] </plan> Patient context: [context]. For each goal, tell me whether it is measurable, patient-centered, realistic for the timeframe, and tied to the stated problem. Rewrite the ones that are not, keeping my clinical intent, so that another nurse could tell objectively whether it was met. For each intervention, check that it is specific enough to act on (what, how often, by whom) and that the rationale connects it to the goal. Output a two-column before/after table plus a short list of interventions that have no matching goal or goals with no supporting intervention.

You are helping structure the teaching portion of a care plan. Patient: [patient]. Learning needs I identified: [learning needs]. For each need, outline: the teaching objective in patient-centered measurable terms, the method suited to this patient's barriers, when in the stay to teach it, who else should be present, and how I will verify understanding using teach-back. Sequence the topics so the most safety-critical comes first and the patient is not overloaded in one session. Flag any need that realistically cannot be met before discharge and should be handed to [follow-up resource]. Present as a table, and keep all patient-facing wording at a 6th-grade reading level.

You are a transactional attorney reviewing a draft agreement for a client. This is a first-pass issue spot to support attorney review, not legal advice. Agreement: <contract> [contract] </contract> Client position: [client position]. Deal context: [deal context]. Client's priorities: [priorities]. Produce a table: Clause reference, What it says in plain English, Risk to our client, Severity (high/medium/low), Suggested revision. Cover at minimum indemnification, limitation of liability, termination, IP ownership, confidentiality, payment terms, assignment, governing law, and dispute resolution. Separately flag: terms that are unusual for this deal type, anything internally inconsistent or undefined, and clauses that are missing entirely. Quote exact clause language. Do not opine on jurisdiction-specific enforceability without noting the need to verify.

Act as a lawyer explaining contract language to a sophisticated but non-legal client. Here is the clause: <clause> [clause] </clause> Context: [context]. Explain: what it means in plain English, what it means practically for the client's business (a concrete scenario, not abstractions), who bears the risk, what would have to happen for this clause to matter, and what a more balanced version would look like. Write at a level a busy executive can read in two minutes. Avoid Latin and defined-term shorthand unless you explain it. End with the one question the client should answer before we decide whether to push back. Note that this is for client discussion, not a legal opinion.

You are reviewing the counterparty's redline against our standard form. Our standard: <standard> [our clause] </standard> Their version: <redline> [their version] </redline> Deal context: [context and leverage]. For each substantive change, report: what shifted, which party it favors, the practical consequence, and whether it is market-standard, aggressive, or unacceptable given our position. Ignore purely stylistic edits but note if a 'stylistic' edit quietly changes meaning. Then give me a negotiation posture: which changes to accept, which to counter (with proposed language), and which are walk-away items. Rank by how much each is worth fighting for. Highlight anything you are uncertain about for attorney verification.

You are a senior associate structuring a research memo for a partner. Do not invent case citations — where authority is needed, describe the type of authority to find and mark it as [CITE NEEDED]. Question presented: [legal question]. Jurisdiction: [jurisdiction]. Facts: [key facts]. Client's goal: [goal]. Produce a memo with: Question Presented · Brief Answer (two sentences, with your confidence level) · Facts · Discussion organized by IRAC for each sub-issue · Counterarguments the other side will make and how we respond · Conclusion and recommended next steps. In the Discussion, be explicit about which parts of the analysis turn on settled law, which turn on the facts, and which are genuinely uncertain. List the specific research I still need to run to close each gap.

Act as a litigation strategist. Before I start researching, help me map the problem. Situation: [facts]. Jurisdiction: [jurisdiction]. Our client is [position]. Opposing party's likely claim or defense: [expected claim]. Break this into an issue tree: the top-level legal questions, the elements each one requires, and the factual questions that determine each element. For every element, note what evidence would prove it and what we currently have or lack. Then rank the issues by how outcome-determinative they are, so I research the dispositive ones first. Flag any threshold issues — jurisdiction, standing, limitations periods, procedural bars — that could end this before the merits. Do not cite specific cases; identify what to look for.

You are drafting from source material I provide. Use only these materials; do not add outside authority. <materials> [materials] </materials> Issue: [issue]. Our position: [position]. Audience: [audience]. Synthesize the materials into a section that states the governing rule, shows how the authorities apply it, applies it to our facts ([facts]), and distinguishes any adverse authority in the materials. Quote sparingly and only where the exact language matters. Where the materials conflict or leave a gap, say so explicitly rather than papering over it. After the draft, list every proposition that needs a pin cite and every place where I should verify the source myself before filing.

You are an attorney writing a status update to a client. Matter: [matter type]. Client sophistication: [client sophistication]. What has happened: [developments]. What it means: [assessment]. What is next: [next steps]. Anything the client must do: [client actions]. Write an email that leads with the bottom line, explains what changed in plain English, states honestly what is still uncertain, and ends with a clearly labeled list of what the client needs to do and by when. Keep it under 250 words. No legalese, no hedging stacked on hedging. Set expectations about timing realistically. Then give me a one-paragraph version for a client who only reads the first three lines.

Act as an experienced attorney communicating an unfavorable development to a client. Situation: [situation]. Impact on the client: [impact]. What we can still do: [options]. My recommendation: [recommendation]. Draft a message that states the news directly in the first two sentences without burying it, explains why it happened factually, lays out the realistic options with their costs and likely outcomes, gives a clear recommendation with reasoning, and offers a call. Do not over-apologize, over-promise a recovery, or hide behind procedural language. Tone: steady, candid, in control. Under 300 words. Then list the three questions this client will most likely ask in response, with a direct answer prepared for each.

You are preparing for a first meeting with a prospective client about [matter type] in [jurisdiction]. What I know so far: [brief facts]. Give me a structured intake question list covering: the factual timeline, the parties and relationships, documents and communications that exist, deadlines and limitations issues, prior counsel or proceedings, the client's actual goal (which may differ from what they asked for), budget and risk tolerance, and conflicts-check information. Order the questions so the ones that could disqualify the matter or reveal urgency come first. For each, add a one-line note on why it matters and what answer would be a red flag. Keep it to a single page I can use live in the meeting.

You are a litigator preparing to depose a witness. Case: [case]. Witness: [witness]. What I need to establish: [goal]. Key documents: [key documents]. Known problems with our theory: [weaknesses]. Build an outline organized by topic, not chronology, moving from background and foundation to the contested facts. For each topic give: the objective, a funnel of questions from open to closed, the document to put in front of the witness, and the specific admission I want on the record. Add a section on anticipated evasions with the follow-up that closes the door. Mark which questions I should not ask because I do not know the answer. Keep questions short and single-fact.

Act as opposing counsel preparing to cross-examine my witness, so I can prepare them. Witness background: [background]. Their testimony: [summary]. Documents they authored or received: [summary]. Known vulnerabilities: [vulnerabilities]. Write the 15 hardest questions the other side will ask, in the order they would ask them to build a trap. For each, note what they are trying to establish and the honest answer that does not concede more than the facts require. Then list the three habits this witness most needs to practice (answering only what is asked, not guessing, correcting mischaracterizations) with a one-line coaching note for each. Do not suggest anything that shades into coaching false testimony.

You are a litigation associate reviewing a deposition transcript. <transcript> [transcript] </transcript> Our claims/defenses: [claims/defenses]. Elements we must prove: [elements]. Produce: (1) a table of admissions helpful to us with page:line cites and why each matters, (2) harmful testimony with page:line and how we might blunt it, (3) inconsistencies within this transcript or with [other evidence], (4) topics the witness would not or could not answer, and (5) follow-up discovery this testimony opens up. Quote exactly and never paraphrase a quote into something stronger than the words support. Flag any passage where the meaning depends on context I have not given you.

You are a staff accountant helping troubleshoot a reconciliation. Account: [account]. Period: [period]. Book balance: [amount]. Statement balance: [amount]. Difference: [amount]. Transactions: <data> [transactions] </data> Work through the usual suspects in order and tell me which the numbers actually support: timing differences, transposition errors (check whether the difference is divisible by 9), duplicated entries, sign errors (difference divisible by 2), missing fees or interest, FX differences, and items posted to the wrong period or account. Output: the most likely explanation with the arithmetic that supports it, the specific transactions to inspect, and the journal entry to correct it if confirmed. Say clearly when the data provided is insufficient to conclude.

Act as a controller documenting a repeatable process. We reconcile [account type] in [system] every [frequency]. Current pain points: [pain points]. Team: [team]. Write a step-by-step procedure a new hire could follow: source documents to pull, matching order, tolerance thresholds and who approves exceptions above them, how to document unmatched items, sign-off and review, and the evidence to retain for audit. Add a short list of the errors this process is specifically designed to catch, and the ones it will not catch (so we know the residual risk). Flag any step where segregation of duties matters. Present as a numbered procedure plus a one-page checklist version.

You are an Excel and spreadsheet expert working with an accountant. My data: [data]. What I need: [desired output]. Give me the formula (Excel [version] / Google Sheets), written to be readable rather than clever, with each part explained in one line. If a helper column makes it simpler and more auditable, use one and say why. Then tell me: how the formula behaves on blanks, text-formatted numbers, duplicates, and dates stored as strings — the four things that break accounting spreadsheets. Include one test row I can use to confirm it works before I trust it on the full file.

You are a controller building a close checklist for a [company type and size] on [accounting basis] using [ERP/system]. Complexities we have: [complexities]. Close deadline: [business day]. Produce a day-by-day checklist from day -2 through sign-off. For each task: owner role, dependency, system, and the evidence produced. Sequence it so nothing waits on a task scheduled after it. Separately list the five tasks that most often cause a close to slip and a mitigation for each. Mark which tasks could be moved earlier into the month to shorten the close, and which controls must not be compressed. Output as a table plus a short critical-path note.

Act as a financial reporting analyst. Here are the numbers: <data> [numbers] </data> Business context: [business context]. Identify the variances that matter — use both dollar and percentage thresholds, and say which threshold you applied. For each, give the plausible drivers based on the context I provided, and separate volume effects from rate or price effects where the data allows. Write the commentary as management would want to read it: one sentence stating the variance, one explaining the cause, one on whether it recurs next period. Flag any variance you cannot explain from the data and list what I would need to pull to explain it. Do not speculate beyond the numbers.

You are a senior accountant. For each situation below, tell me the correct treatment under [standard] and the journal entry. <situations> [situations] </situations> Entity: [entity]. For each, output: the recognition principle at stake, the entry (account, debit/credit, amount or formula), the reversal treatment next period, and the support to retain. Where the answer depends on a fact I have not given (contract terms, transfer of control, materiality), state the question rather than assuming. Note anything that would draw an auditor's attention. This is drafting support — final treatment is my professional judgment.

You are an accountant writing to a small business owner who does not read financial statements. Here are their numbers: <financials> [financials] </financials> Business: [business]. Their concerns: [concerns]. Write a one-page summary: how the business did this period in one sentence, the three numbers that matter most for this business and what each means, what changed versus last period and why, cash position and runway in plain terms, and two specific things to watch next quarter. Use everyday language and concrete comparisons rather than ratios alone. No jargon without a plain-English gloss. Be honest about anything concerning — do not soften it. End with the two questions I should ask them.

Act as an advisory-minded accountant reviewing a client's results. <financials> [financials] </financials> Industry: [industry]. Size: [size]. Owner's goal: [goal]. Go beyond describing the numbers. Identify trends across periods, the metrics most relevant to this industry and where they sit, and the two or three levers with the largest effect on the owner's stated goal. For each recommendation give: the observation with figures, the action, the expected financial effect, and the risk. Rank by impact per unit of effort. Explicitly separate what the data shows from what you are inferring, and list the information you would need to firm up the weaker inferences.

You are preparing me to walk a client through their [period] results. Client: [client]. Results summary: [results]. Sensitive topics: [sensitive topics]. Give me a meeting outline with time allocations, the three points that must land, and a plain-language way to explain [hardest concept]. Then anticipate: the five questions this client will ask, with a direct answer for each, and how to raise [sensitive topic] without triggering defensiveness. Include one question I should ask them about their plans that could open advisory work. Keep the whole thing to one page I can hold during the call.

You are a tax preparer building an information request list. Client profile: [client profile]. Tax year: [year]. What changed since last year: [changes]. Produce a checklist of documents and information to request, grouped by category (income, expenses, assets, credits, prior-year carryforwards, entity items). For each item, note why it is needed and which situations make it required versus optional. Highlight the items specific to this client's situation that a generic checklist would miss, and the three most common items clients forget. Format it as a client-facing request list in plain language plus an internal version with the technical reason. Note that positions require verification against current law.

Act as a tax adviser reviewing a client's situation for planning ideas and exposure. This is analysis support, not a filing position — everything must be verified against current law for [jurisdiction and tax year]. Client facts: <facts> [client facts] </facts> List: (1) planning opportunities the facts suggest, each with the mechanism, the rough benefit, and the conditions that must hold; (2) positions in the current facts that carry audit risk, with what documentation would support them; (3) deadlines and elections that could be missed. Rank by dollar impact. For each item, state exactly what I must verify before relying on it, and flag anything where the law has been changing.

You are a tax professional explaining an outcome to a client who is surprised by it. Situation: [situation]. Drivers: [drivers]. Write an explanation the client can follow: the bottom line first, then each driver quantified in plain language, then what (if anything) could have changed the result, then what to do differently next year with specific actions and deadlines. Avoid code sections and jargon; use dollars and simple cause-and-effect. Do not blame the client, and do not imply an outcome was avoidable if it was not. Under 300 words, plus a short bulleted 'next year' action list. Note that figures are based on the information provided.

You are a B2B sales rep who writes short, human emails — no 'I hope this email finds you well'. Prospect: [prospect]. Trigger I noticed: [trigger]. What we sell: [what we sell]. Proof: [proof]. Ask: [ask]. Write an email under 90 words: a first line that could only be written to this person, one sentence connecting the trigger to a problem people in their role have, one line of proof, and a low-friction ask. No feature lists, no adjectives about ourselves. Then give two subject lines under 40 characters and a one-line note on the assumption this email is betting on.

Act as a sales development leader. Build a [number]-touch sequence over [days] for [persona] at [company type], mixing email, LinkedIn, and phone. What we sell: [product]. Problem it solves: [problem]. Why now: [why now]. Competitors they may use: [competitors]. Each touch must add something new — a different angle, proof point, or objection handled — never 'just bumping this'. Specify for each: day, channel, angle, full copy (email under 90 words, LinkedIn under 50, voicemail under 20 seconds), and the reply this touch is designed to provoke. Include a genuine break-up message at the end. Then list the two touches you would cut first if reply data says the sequence is too long.

You are a cold email specialist. Here is my email, which has a [reply rate] reply rate to [persona]: <email> [email] </email> Diagnose it against the reasons cold emails fail: it is about us not them, the relevance is generic, the ask is too big, it is too long, the value claim is unbelievable, or the first line reads like a template. Name the primary failure with evidence from the text. Then rewrite it three ways — one that leads with the prospect's problem, one that leads with a customer result, one that leads with a genuine question — each under 90 words. For each version say who it would work best on and what it risks. Finish with the single line in my original that is worth keeping.

You are a sales coach preparing me for a discovery call. Prospect: [prospect]. Person: [person]. How they came in: [how they came in]. What we sell: [product]. What we need to qualify: [qualify]. Write 12 questions grouped as: current state, the cost of the problem in their own numbers, what they have already tried, decision process and who else is involved, and what success would look like in 90 days. Make them open and specific to this persona — no questions answerable with yes or no, and none I could answer myself from their website. For each, add the follow-up to ask if the answer is vague. End with the two questions that most often reveal a deal is not real.

Act as a sales researcher. Here is what I have on the account: <info> [account info] </info> We sell [product] to [persona] to solve [problem]. Build a one-page pre-call brief: what this company does in one sentence, the business priority the evidence suggests they are chasing, likely pressures on the person I am meeting, the language they use for the problem we solve, what they probably use today, and the two most credible hypotheses for why they would care about us now. Separate observed facts from inference and mark each. End with three specific questions that show I did the reading, and one thing I should avoid assuming.

You are a sales manager reviewing my call notes. <notes> [notes] </notes> Deal stage: [deal stage]. Our sales process requires: [qualification criteria]. Produce: a summary of what the prospect actually said (using their words for the problem and the impact), what we now know versus still do not know against each qualification criterion, the strength of the pain on a 1-5 scale with evidence, risks to the deal, and the specific next step with a date. Then draft a recap email to the prospect under 150 words that mirrors their language, confirms the next step, and includes one thing of value. Flag anything in my notes that is my interpretation rather than something the prospect said.

You are a sales strategist. Deal context: [deal context]. Objection I am hearing: [objection]. First, tell me what this objection usually means underneath — a real constraint, a priority problem, a trust gap, or a stall. Give the diagnostic question that tells me which. Then give three responses calibrated to each underlying cause. Each one: acknowledge without agreeing, reframe with a question or a fact, and propose a concrete next step. Keep them conversational, under 60 words each, with no manipulation tactics. Finish with what I should do if the objection is legitimate and we genuinely are not the right fit.

Act as [title] at [company type] evaluating [product]. Your situation: [situation]. You are [persona] and you interrupt when answers get long. Roleplay a [length]-minute [call type] with me. Stay in character, respond only as the prospect, and push back the way a real buyer would — vague answers, competing priorities, and the objection [specific objection] at some point. Do not make it easy, and do not agree to a meeting unless I have earned it. When I type 'COACH', break character and tell me what I just did well, what I missed, and the better line. Start with your first response to my opening.

You are a negotiation coach for a [product] deal worth [amount] with [prospect situation]. They are pushing for [discount or concession]. Our constraints: [constraints]. Give me: what their request likely signals about the deal's real status, the questions to ask before responding at all, a set of trades where every concession we give buys something back, the specific language for holding price without sounding rigid, and the walk-away point. Write the three lines I should actually say, in order. Then tell me what to do if they go silent for a week. Keep it practical — no theory.

You are a sales rep who never sends an empty follow-up. Context: last contact was [when] and we discussed [topic]. They committed to [their commitment]. Since then: [since then]. Deal value: [deal value]. Their stated priority: [their priority]. Write three follow-up options, each under 80 words and each carrying something new: one that shares a relevant resource or customer example, one that surfaces a risk or deadline they should know about, and one that asks directly whether this has fallen down the priority list and gives them an easy out. For each, say when to use it. Then tell me at what point I should stop following up and what the final message should say.

Act as an enterprise seller writing a proposal. Everything must come from what the buyer told us. What they said: <discovery> [discovery notes] </discovery> Our solution and pricing: [details]. Write a proposal structured as: their situation in their language, the cost of leaving it as-is (using their numbers), the outcome they defined as success, how we deliver it (mapped point by point to their criteria), implementation timeline, investment, and the specific next step with a date. Keep it under two pages. Every claim tied to something they said or a proof point we have. Flag any section where discovery was too thin to write it properly.

You are a sales manager doing deal inspection. Here is the deal: <deal> [deal] </deal> Assess honestly whether this deal is real. Check: Do we have access to the person who can say yes? Is there a compelling reason to act by their date, or is the date ours? Do we know their decision process and competing priorities? Has anyone told us no in a polite way? Give a forecast confidence with reasoning, the three riskiest gaps, and one specific action for each that would resolve it within two weeks. If this deal should be pushed out or disqualified, say so directly.

You are an HR business partner writing a job posting for [role] at [company]. Team context: [team context]. Must-have skills: [list]. Nice-to-have: [list]. Location/work model: [work model]. Range: [range]. Write: a two-sentence hook about the actual work, what the person will do in their first 90 days, ongoing responsibilities as outcomes rather than task lists, genuine requirements (cut anything that is really a preference), what we offer, and how to apply. Use inclusive, concrete language — no 'rockstar', no 'wear many hats' unless we mean it, no degree requirement unless the job truly needs one. Then flag any requirement that will narrow the candidate pool more than it improves quality.

Act as a hiring adviser reviewing a job posting for effectiveness and fairness. <posting> [posting] </posting> Role level: [role level]. Market: [market]. Problem we are having: [problem]. Analyze: which requirements are actually necessary versus inherited from a template, language that skews the applicant pool (gendered wording, unnecessary credentials, culture-fit signals), how quickly a qualified candidate can tell whether this job is for them, and whether the posting sells the role or just lists demands. Give a marked-up version with specific edits and a short rationale for each. Then name the single change most likely to fix the stated problem, and one thing in the posting worth keeping exactly as it is.

You are an org design consultant. Before we post [role], help scope it properly. Context: [context]. Ask the questions we should have answered: What outcomes will this person own that nobody owns now? What decisions can they make alone? Where does this role overlap with [existing roles]? What would success look like at 6 and 12 months? Is this one role or two? Then propose a role scope with: three to five accountabilities, the seniority level the accountabilities imply, the skills that genuinely predict success, and the level of role we can realistically attract at [budget]. Flag if the scope as described is unrealistic for one person.

You are an HR partner helping a manager write a fair performance review. Employee: [employee]. Review period: [period]. Goals set: [goals set]. Evidence I have: [evidence]. Overall assessment: [overall assessment]. Write the review using Situation-Behavior-Impact for every point. Ground each statement in a specific example from my evidence — never a personality trait. Cover: results against goals, how they worked with others, growth since last period, the two highest-impact development areas, and goals for next period that are specific and measurable. Keep the tone direct and respectful, consistent with the rating (no 'exceeds expectations' language attached to a 'meets' rating). Flag any claim my evidence does not actually support.

Act as an experienced people leader coaching me. I need to tell [employee] that [difficult message]. Evidence: [evidence]. Their likely reaction: [likely reaction]. Their strengths: [strengths]. Give me: an opening that delivers the message clearly in the first 30 seconds, the evidence framed as impact rather than judgment, a path forward with specific and observable criteria, and the boundary between explaining and negotiating. Then script responses to the three hardest things they might say ('this is the first I'm hearing of it', 'so-and-so does the same thing', an emotional reaction). Note what I must document and what I should not say. Keep it under one page.

You are facilitating review calibration. Here are draft ratings and justifications for a team: <reviews> [reviews] </reviews> Our rating definitions: [rating definitions]. Check for: ratings whose justification does not match the definition, inconsistent standards across people at the same level, recency bias (evidence clustered in the last month), justifications built on personality rather than outcomes, and language patterns that differ systematically between groups. Output a table: person label, issue found, what to fix. Then list the questions I should put to managers in the calibration session, ordered by how much the answer could change a rating. Note where I would need more evidence to judge.

You are an HR policy writer. Draft a [policy topic] policy for a [company size] company in [industry], operating in [locations]. Our culture: [culture]. What prompted this: [reason]. Practical constraints: [constraints]. Structure it as: purpose in one sentence, who it applies to, the actual rules stated as what people should do (not a list of prohibitions), the process for requests or exceptions with who decides, what happens when the policy is not followed, and where to ask questions. Write in plain language and second person. Every rule must be enforceable and unambiguous. Mark any section that needs employment-counsel review for [jurisdictions], and flag any rule that will be widely ignored because it is impractical — those create more risk than they remove.

Act as an internal communications specialist. Here is an existing policy: <policy> [policy] </policy> Audience: [audience]. Complaint we get: [complaint]. Rewrite it at a grade 8 reading level without changing a single obligation or entitlement. Replace legalese with plain words, convert dense paragraphs into scannable structure, lead each section with the answer to the question employees actually have, and add a short FAQ addressing the three most common misunderstandings. Provide the rewrite plus a side-by-side table of any wording change that a lawyer should confirm is substantively identical. Do not soften or strengthen any requirement on your own.

You are writing an internal announcement about a policy change. What is changing: [change]. Why: [reason]. Effective: [date]. Who it affects and how: [impact by group]. What we considered and rejected: [alternatives]. Expected reaction: [expected reaction]. Write an announcement that states the change and the date up front, gives the honest reason without corporate euphemism, says plainly what each affected group must do differently, acknowledges the downside for those who lose something, and names where to ask questions and by when. Under 400 words. No 'exciting news' framing for something that is not good news. Then give me an FAQ of the six hardest questions with straight answers, including the ones we would rather not be asked.

You are a people operations lead. Build an onboarding plan for a new [role] joining [team] at a [company size] [industry] company. They report to [role]. What they must be able to do independently by day 90: [outcomes]. Systems and access needed: [systems and access]. Team's current state: [team state]. Produce a plan by phase: day 1, week 1, days 8-30, 31-60, 61-90. For each, list learning goals, the specific work they will do, who they meet and why, and the check-in that confirms they are on track. Make day 1 about belonging and access, not policy videos. Assign an owner to each item. Flag anything that will slip if the manager is busy, and what the backup is.

Act as an HR partner writing a checklist for a manager who is bad at onboarding but well-meaning. New hire: [new hire]. Team context: [context]. Write a checklist covering: before day 1 (access, equipment, announcement, buddy assignment, first-week calendar), day 1, week 1, and each week through week 6. Every item must be a concrete action with an owner and a deadline — not 'make them feel welcome'. Include the specific questions to ask in the week 1, week 4, and week 8 check-ins that actually surface problems early. Add three signals that onboarding is going badly and what to do about each. Keep it to one page, formatted as checkboxes.

You are designing onboarding for a fully remote [role] joining a team in [time zones], where most of the team is [team setup]. Known risks: [known risks]. Design the first 30 days to solve the two things remote onboarding fails at: building relationships without hallway time, and learning the unwritten context that never gets documented. Specify: the structured social touchpoints and who initiates them, how to make implicit knowledge explicit (which docs to write, which recordings to make), async-friendly work for their first tasks, meeting overlap expectations, and the escalation path when they are stuck at 9pm their time. Flag anything that requires the team to change its habits, since that is where these plans usually fail.

You are a senior product manager writing a PRD for engineering and design. Feature: [feature]. Problem it solves: [problem]. Users affected: [users affected]. Business goal and metric: [goal]. Constraints: [constraints]. What we already know: [what we know]. Write: problem statement · why now · target user and their current workaround · goals and explicit non-goals · user stories with acceptance criteria · key flows described step by step · edge cases and error states · success metrics with baseline and target · open questions with owners · what we are deliberately not building in v1. Keep it under two pages. Mark every assumption. If the problem statement is too vague to build from, say so before writing the rest.

Act as a skeptical head of product reviewing this idea before it gets staffed. <idea> [idea] </idea> Evidence we have: [evidence]. Cost estimate: [effort]. Alternative uses of that effort: [alternatives]. Challenge it: Is this a real problem or a requested solution? How many users actually hit it and how often? What do they do today, and is that workaround good enough? What breaks if we build it (support load, complexity, edge cases)? What is the cheapest experiment that would tell us we are wrong? Give a recommendation — build, shrink, test first, or drop — with the reasoning and the one piece of evidence that would change your mind.

You are a product manager writing acceptance criteria. Feature: [feature]. Intended behavior: [behavior]. Users: [users]. Platforms: [platforms]. Known constraints: [constraints]. Write the criteria in Given/When/Then form, covering the happy path, every input validation case, empty and loading states, error states with the exact user-facing message, permissions differences by user type, and what happens on slow or failed network. For each criterion, make it objectively testable — a QA engineer must be able to mark it pass or fail without asking me. Then list the decisions I have not made yet that are blocking a complete spec, phrased as concrete questions with the options. Do not fill those gaps with your own assumptions.

You are an agile product owner. Break this epic into user stories. Epic: [epic]. Context: [context]. Team: [team]. Sprint length: [sprint length]. Produce stories that each deliver observable user value and could ship independently. For each: the story in 'As a / I want / so that' form, acceptance criteria as bullet points, dependencies, and a rough size (S/M/L) with the reason. Sequence them so the riskiest assumption is tested first and something usable ships earliest. Explicitly separate stories that are true slices of value from technical enablers, and say which enablers are genuinely required first rather than merely convenient. Flag any story that is really three stories wearing a trench coat.

Act as a product owner refining a ticket that engineering has pushed back on. <ticket> [ticket] </ticket> What I actually want: [intent]. Context: [context]. Rewrite it to be ready for sprint planning: clear user and outcome, explicit scope boundaries, acceptance criteria, the states and edge cases to handle, design or copy needed, and how we will verify it worked. Before the rewrite, list the questions the ticket leaves unanswered that only I can answer, and mark which of them block estimation versus which can be resolved during implementation. Do not invent product decisions — where I must decide, present the options and a recommendation.

You are preparing me to run a refinement session. Here is the backlog slice we will discuss: <backlog> [backlog] </backlog> Team: [team composition]. Sprint goal candidate: [sprint goal]. Known constraints: [constraints]. For each item, assess readiness: is the value clear, are the criteria testable, are dependencies identified, is it small enough for one sprint? Mark each ready / needs work / not ready with the specific gap. Then give me the three questions per not-ready item to ask in the session, the order to discuss items so dependencies surface early, and a realistic sprint scope given the team's capacity. Call out anything that looks like scope we accepted without deciding.

You are a product leader applying [framework] rigorously. Here are the candidate initiatives: <initiatives> [initiatives] </initiatives> Company goal this quarter: [goal and metric]. Team capacity: [team capacity]. Constraints: [constraints]. Score each on the framework's dimensions, showing the number and the reasoning behind it — especially confidence, since that is where these scores usually lie. Then rank them. After the table, do the part the framework misses: which items are strategically necessary despite a poor score, which high scorers are actually redundant, and what the ranking assumes about [key assumption]. Recommend the quarter's set with the reasoning, and name what we are explicitly saying no to.

Act as a product leader helping me decline a request without damaging the relationship. Request: [request]. Who asked: [requester]. What is really behind it: [underlying need]. Why we are not doing it: [reason]. What we are doing instead: [priorities]. Draft a response that shows I understood the underlying problem, gives the real reason (capacity, strategy, evidence) rather than a vague 'not on the roadmap', names what would have to change for this to move up, and offers the nearest thing we can do — including a manual workaround if one exists. Under 200 words, respectful and specific. Then give me the one-line version for a hallway conversation and a note on what to document so this does not resurface monthly.

You are a VP of Product presenting a roadmap. Audience: [audience]. Timeframe: [timeframe]. Planned work: [planned work]. Company strategy: [company strategy]. Last period's results: [last period]. Write a narrative, not a feature list: the two or three bets we are making and why, what each is expected to move with a target number, what we learned last period that changed our thinking, what we are deprioritizing and the cost of that, and the risks with mitigations. Structure as a one-page memo plus a slide outline. Anticipate the four questions executives will ask, including 'why is this taking so long', with a direct answer for each.

You are a user researcher synthesizing feedback. Here is the raw data: <feedback> [feedback] </feedback> Product: [product]. Segment: [segment]. Question I am trying to answer: [question]. Cluster into themes by the underlying problem, not the surface wording. For each theme: a name in the customers' own language, how many mentions, representative verbatim quotes, the user segment it concentrates in, whether it is a bug, a usability failure, a missing capability, or an expectation mismatch, and the severity of its impact. Rank themes by frequency times severity. Then state clearly what this data cannot tell us — sampling bias, who is not represented, and which themes rest on one or two comments.

Act as a research analyst. Here is an interview transcript: <transcript> [transcript] </transcript> Research goal: [research goal]. Participant: [participant]. Extract: the participant's actual workflow step by step, where they hesitated or worked around something, the moments of visible frustration or delight with the quote, their stated needs versus their demonstrated behavior (call out where these differ), the vocabulary they use for the domain, and anything that contradicts our current assumptions. Separate what they said from what I might be tempted to conclude. Flag leading questions in the transcript that may have shaped answers. Finish with the three follow-up questions I should ask the next participant to test the strongest emerging pattern.

You are a product manager translating research into decisions. Synthesized findings: <findings> [findings] </findings> Business goal: [business goal]. Capacity: [engineering capacity]. Current roadmap: [current roadmap]. For each finding, decide: fix now, schedule, investigate further, or accept and document. Justify each with the affected user volume, the severity, and the cost of the fix. Separately identify: quick fixes worth doing this sprint, findings that indicate a deeper structural problem rather than a surface bug, and findings that conflict with our current roadmap (say which should win). End with what we should tell the customers who raised these, and the metric that would show whether our response worked.

You are a senior analytics engineer writing [dialect] SQL. Here is my schema: <schema> [schema] </schema> Known data quirks: [data quirks]. Business question: [question]. Before writing SQL, restate the question as a precise definition — what counts as the entity, the time window, and any exclusions. If my question is ambiguous, state the interpretation you are using. Then write the query with CTEs rather than nested subqueries, a comment per CTE, and explicit handling of nulls and duplicates. Finish with a sanity check I can run to confirm the result is plausible.

Act as a database performance specialist. Here is my query and what it does wrong: <query> [query] </query> Problem: [problem]. Dialect: [dialect]. Table sizes: [table sizes]. Indexes: [indexes]. Query plan if available: [query plan]. If the results are wrong, trace the logic step by step and find where rows are being duplicated, dropped by an inner join, or filtered by a WHERE clause that should be in the ON clause. If it is slow, identify the specific operation causing it — full scan, spilled join, function on an indexed column, or an unnecessary DISTINCT. Give the corrected query, a diff-style explanation of each change, and how to verify the new version returns the same intended rows.

You are a data engineer explaining inherited code. Here is a query I need to modify: <query> [query] </query> Schema context: [schema context]. What I need to change: [goal]. Explain: what business question this query answers, what each CTE or subquery contributes, the grain of the output (one row per what?), every filter and what it excludes, and any join that could silently drop or duplicate rows. Then tell me specifically where to make my change and what else it would affect. Flag anything in the query that looks like a bug or a stale assumption — hardcoded dates, magic IDs, filters that no longer make sense — but do not change them without telling me.

You are a senior data analyst. Business question: [question]. Decision it informs: [decision]. Data available: [data available]. Deadline: [deadline]. Design the analysis before any code: the precise metric definition and why that definition and not a neighboring one, the population and exclusions, the comparison or baseline that makes the number meaningful, the cuts worth examining, and the confounders that could make a naive answer wrong. Then list the data quality checks to run first, and state what result would mean the analysis cannot answer the question honestly. Finish with the one chart or table that would answer the question if everything checks out. Keep it under 400 words.

Act as a data scientist writing [language] for a first look at a new dataset. Dataset: [dataset]. Context: [context]. Goal: [goal]. Write a script that: profiles missingness and its pattern, checks the grain is what I think it is (duplicate keys), summarizes distributions including outliers, checks date ranges and gaps, cross-tabs the key categorical fields, and flags values that are impossible given the domain. Keep it readable, no unnecessary libraries. After the code, list the five specific things I should look at in the output and what each would imply about whether this data can answer my question.

You are a statistician reviewing my analysis for overclaiming. Here are my results: <results> [results] </results> Method used: [method]. Sample: [sample]. Question: [question]. Conclusion I want to draw: [conclusion]. Tell me whether the data supports that conclusion. Check specifically: sample size and whether differences are within noise, selection bias in how the data was collected, confounders that could explain the pattern, whether this is correlation being described as cause, multiple comparisons, and whether the effect size matters practically even if it is statistically detectable. Give me: the conclusion the data actually supports, the caveats that must accompany it, and the additional data or test that would strengthen it.

You are an analytics lead designing a dashboard for [audience]. Decisions they make: [decisions]. Data available: [data available]. Tool: [tool]. Start from the decisions, not the data. For each decision, specify the one metric that informs it, its precise definition, the comparison that makes it interpretable (target, prior period, segment), the chart type and why, and the threshold at which the viewer should act. Then lay out the dashboard: what goes above the fold, what belongs in a drill-down, and what should not be on it at all. Name the metrics people will ask for that you are deliberately excluding and why. Include the refresh cadence and who owns each definition.

Act as an analytics engineer resolving a metric definition dispute. Metric: [metric name]. How different teams currently define it: [current definitions]. Underlying data: [underlying data]. What the metric is used to decide: [use]. Write a single canonical definition covering: the numerator and denominator in words, the entity and time grain, inclusions and exclusions (test accounts, internal users, refunds, deleted records), how late-arriving data is handled, and the timezone. Then show the SQL that implements it, list the edge cases where reasonable people would disagree with a recommendation for each, and quantify roughly how much the number changes between the competing definitions so people understand what is at stake.

You are a data visualization reviewer. Here is a description of my dashboard: <dashboard> [dashboard] </dashboard> Audience: [audience]. What they should do with it: [intended use]. Complaint: [complaint]. Assess: does each chart support a decision or is it there because the data existed? Is the chart type right for the comparison being made? Can a viewer tell in five seconds whether things are good or bad? Are there missing baselines, misleading axes, or metrics that move together and are redundant? Give a revised layout, the charts to cut, and the one addition most likely to fix the stated complaint. Explain each change in one line.

You are a analytics director writing for executives who read the first three lines. Analysis: <analysis> [analysis] </analysis> Audience: [audience]. Decision at stake: [decision]. My recommendation: [recommendation]. Write it top-down: the recommendation and the number that justifies it first, then the three supporting findings each in one sentence with the figure, then the caveats that could change the conclusion, then what happens next and who owns it. No methodology in the body — put it in an appendix note. Translate every metric into business impact (dollars, customers, hours). Under 300 words. Then give me the single sentence I would say if I had one line in a meeting.

Act as a skeptical executive who has been burned by bad analysis. Here is what I am about to present: <presentation> [presentation] </presentation> Method: [method]. Audience: [audience]. Ask me the eight hardest questions, in the order they would come up: about the data source, the definition of the metric, what is not in the sample, alternative explanations, why the number differs from [another team's number], the cost of acting on this, and what would happen if we did nothing. For each, tell me whether I can answer it from what I have, and if not, what I need to prepare. Then name the one weakness in this analysis that I should disclose myself before someone finds it.

You are translating a technical result for people who do not do statistics. Result: [result]. Audience: [audience]. What they need to do with it: [action]. Explain it in plain language with a concrete analogy from their domain, state what it means for their decision, and be explicit about the uncertainty in terms they can act on ('between X and Y, most likely around Z') rather than statistical vocabulary. Avoid p-values, significance language, and model jargon entirely. Do not oversimplify to the point of being wrong — if the honest answer is 'we do not know yet', say that and explain what would change it. Under 200 words.

You are a conversion copywriter. Product: [product]. Who it is for: [audience]. The problem they have today and how they solve it now: [current workaround]. What makes us genuinely different: [differentiator]. Proof: [proof]. Objections: [objections]. Awareness level: [awareness level]. Voice: [voice]. Write the full page: headline plus three alternatives, subhead, hero paragraph, three benefit sections each led by an outcome and backed by a feature, social proof placement, objection-handling section, FAQ of five real questions, and the CTA with its supporting microcopy. Write for the awareness level given — do not explain the problem to people who already feel it. No superlatives we cannot prove. Mark every claim that needs verification.

Act as a direct-response copywriter. Current headline: [headline]. Product: [product]. Audience: [audience]. The single most valuable outcome we deliver: [outcome]. What makes it believable: [proof]. Bounce rate or problem: [problem]. Diagnose the current headline first: is it about us instead of them, is it vague, is it a category label rather than a promise, or is it a promise nobody believes? Then write 10 alternatives across distinct angles — outcome, specific number, the problem stated in their words, contrarian, question, before/after — each under 12 words, each paired with a subhead that makes it credible. Rank your top three with the reasoning, and say what each one assumes about the visitor's awareness level.

You are a copy editor who converts features into outcomes. Here is our current copy: <copy> [current copy] </copy> Audience: [audience]. What they actually want: [desired outcome]. Voice guidelines: [voice]. For each feature mentioned, identify the outcome the reader gets and rewrite the line to lead with that outcome, keeping the feature as the reason to believe. Cut every sentence that could appear on a competitor's site unchanged. Return a two-column table of before and after, then the full rewritten copy. Aim to cut total word count by at least 30% without losing information. Flag any feature where I have not given you enough to identify a real user outcome — those probably do not belong on the page.

You are a performance copywriter. Platform: [platform and placement]. Character limits: [limits]. Product: [product]. Audience and their state of mind when they see this: [context]. Offer: [offer]. Proof: [proof]. Landing page promise: [page promise]. Write [number] ad variations organized as a test matrix: hold everything constant except one variable per set — hook angle, proof type, and CTA. Label each with the variable it tests. Every ad must match the landing page promise; no bait. Include primary text, headline, and description within the limits. Then recommend which three to launch first, the metric that decides each test, and the minimum spend or impressions needed before reading results.

Act as a social ad specialist. Product: [product]. Audience: [audience]. The tension or frustration they feel: [pain]. What we can prove: [proof]. Brand voice and what we would never say: [voice and limits]. Write 15 opening lines designed to stop a scroll, spread across these types: a specific number, a common belief contradicted, a moment of recognition ('if you have ever...'), a direct callout of the audience, a short story opener, and a plain useful fact. Each under 12 words. No clickbait we cannot pay off, no fake urgency, no 'attention [audience]!'. Then mark the three you would test first and the one most likely to attract the wrong audience — a cheap click that never converts.

You are a paid media copy analyst. Ads: <ads> [ad copy] </ads> Landing page: [landing page]. Offer: [offer]. Audience targeting: [targeting]. Find the message-match break: does the ad promise something the page does not deliver, does it attract a broader audience than the offer suits, is the CTA implying less commitment than the page asks for, or is the ad selling a different benefit than the page? Quote the specific ad line and page line that conflict. Then rewrite the two worst offenders so ad and page tell one continuous story, and state what you would change on the page if the ad is actually the good half.

You are an ecommerce copywriter for [brand], selling to [audience]. Product: [product]. Who it is for and the moment they use it: [use context]. Compared with alternatives it is better at: [differentiator]. Common pre-purchase questions and returns reasons: [questions and returns]. Voice: [voice]. Write: a one-line hook, a short paragraph that puts the product in the buyer's actual situation, five bullet points where each specification is paired with what it means for the user, sizing or fit guidance if relevant, and a closing line. Answer the known pre-purchase questions inside the copy so buyers do not have to ask. Do not use 'premium', 'luxurious', or 'game-changing' without a concrete detail that earns it.

Act as a catalog copywriter. Here is my product data: <data> [product data] </data> Brand voice: [voice]. Audience: [audience]. Length target: [word count]. Write a description for each, using a consistent structure so the catalog reads as one brand, but varying the opening line and the emphasis so they do not read as templated. For each, lead with the attribute that most differentiates this SKU from its siblings. Output as a table: SKU, description, and the one attribute you led with. Flag any product where the data provided is too thin to write honest copy, and say exactly which attribute I need to supply.

You are an SEO-aware copywriter. Current description: <copy> [description] </copy> Target keyword and close variants: [keywords]. Search intent behind that keyword: [search intent]. Competitor pages ranking above us say: [competitor summary]. Rewrite it so the keyword and its natural variants appear where they belong — title, first sentence, one subhead, image alt suggestion — without a single sentence that reads as written for a crawler. Cover the subtopics the intent implies that our current copy omits. Provide: the rewritten copy, a suggested title tag under 60 characters, a meta description under 155, and a short list of the questions searchers ask that we should add as an FAQ. Do not stuff keywords or invent specifications.

You are a brand strategist. Here are samples of our best writing — the pieces that sound most like us: <samples> [writing samples] </samples> And here is something that sounds wrong for us: [off-brand sample]. Extract the voice: three to four defining traits, each with a one-line definition, what it means we do, and what it means we never do. Note our sentence rhythm, vocabulary level, use of humor, how we handle bad news, and how we refer to the reader and ourselves. Build a do/don't table with real examples pulled from the samples, not invented ones. Then write the same 40-word message three ways — on-voice, slightly off, and clearly off — so the team can hear the difference.

Act as a brand editor. Voice guidelines: <voice> [voice guide] </voice> Copy to fix: <copy> [copy to fix] </copy> Context: [context]. Rewrite it in our voice without changing the facts, the offer, or any legal wording. Then show a short table of the specific changes: the original phrase, the replacement, and which voice trait drove it. Where the source copy contains something our voice cannot say honestly — an overclaim, a tone that does not fit the situation — flag it rather than smoothing it over. If part of the copy is already on-voice, leave it and say so instead of rewriting for the sake of it.

You are a content lead adapting a single message for different contexts. Core message: [message]. Audience: [audience]. Brand voice: [voice]. Occasion: [occasion]. Write it for: a website section, an email, a short social post, an in-product notification (under 20 words), a support macro, and a sales one-liner. Keep the claim and tone consistent while respecting how people read each channel — the in-product version should be scannable in a second; the email can earn a little more room. Do not simply truncate the long version for the short ones. After the set, note any place where the message has to change substantively (not just in length) to be honest in that context, and why.

You are a senior support specialist known for calming situations without grovelling. Customer message: <message> [customer message] </message> What actually happened on our side: [facts]. What we can do: [options and limits]. What we cannot do: [limits]. Policy constraints: [policy]. Customer value/history: [context]. Write a reply that acknowledges the specific frustration (naming the actual inconvenience, not 'we understand your frustration'), states plainly what happened without excuses or blame, gives the concrete resolution with a timeline, and offers one thing that shows we mean it. Do not apologize more than once, do not promise anything outside the stated limits, and do not use passive voice to dodge responsibility. Under 180 words. Then give me a two-line version for chat.

Act as a support lead. The customer is asking for [request]. We cannot do it because [real reason]. What we can offer instead: [alternatives]. Customer context: [customer context]. Write a response that gives the answer in the first two sentences, explains the real reason in plain terms without hiding behind 'policy', offers the best available alternative concretely, and tells them what would have to change for the answer to be different (including how to request it, if there is a path). No false hope, no 'I'll see what I can do' if I cannot. Keep it warm but final. Under 150 words. Then draft the internal note for the ticket recording what was requested and denied.

You are writing to a customer we genuinely let down. What happened: [the failure]. Impact on them: [concrete impact]. What we have fixed: [fix]. What we are doing so it does not recur: [systemic change]. What we can offer: [remedy]. Write an apology that names the specific failure, takes responsibility without hedging ('we did X' not 'X occurred'), acknowledges the real cost to them, states the fix and the prevention, and makes the remedy easy to accept with no hoops. Avoid corporate softeners: no 'we sincerely regret any inconvenience'. Under 200 words. Then tell me what to escalate internally and what this customer will likely need next time they contact us.

You are a support content writer. Topic: [topic]. Product context: [product context]. User skill level: [level]. Where users get stuck: [friction points]. Prerequisites: [prerequisites]. Write: a title phrased the way a user would search for it, a one-line summary of what they will accomplish, prerequisites, numbered steps with the exact UI labels in bold, what they should see after each critical step, and a troubleshooting section for the known friction points. One action per step. No screenshots described in prose — mark [screenshot] where one is needed. End with related tasks. Keep it under 500 words and at a grade 8 reading level.

Act as a knowledge base manager. Here are several tickets about the same issue: <tickets> [tickets] </tickets> Identify the actual underlying question customers are asking (which is often not the one they typed), the variations of wording they use, and the point in their workflow where they get stuck. Then write one article that resolves all variants: title using their words, a short 'if you are seeing X, this is why' opener, the fix, and the two adjacent situations that look similar but need a different fix. Finally, tell me whether this should be an article at all — if the root cause is a product or UI problem, say so and describe the change that would eliminate these tickets.

You are a documentation reviewer. Article: <article> [article] </article> Problem: customers read this and still contact us, saying [what they say]. Their goal: [goal]. Diagnose why it fails: is it answering a different question than they are asking, is the answer buried below the fold, does it assume knowledge they lack, does it describe the feature rather than the task, or is the failing step missing entirely? Rewrite it with the fix. Show the specific structural changes — what moves up, what gets cut, what gets added. Then suggest the title and first sentence most likely to convince a frustrated user they are in the right place, and one product change that would matter more than any article.

You are a support operations manager. Our top ticket drivers this quarter: <data> [ticket drivers] </data> Brand voice: [voice]. Tone rules: [tone rules]. Channels: [channels]. For each of the top drivers, write a macro with: a clear internal name, the trigger conditions for when an agent should use it, the customer-facing text with [placeholders] for the details that must be personalized, and a note on what the agent must verify before sending. Keep each macro short enough that agents will not rewrite it. Mark the ones that should never be sent without personalization. Then flag any driver where a macro is the wrong answer because the underlying problem should be fixed in the product.

Act as a support quality lead. Here is a macro agents currently use: <macro> [macro] </macro> Customer complaint about it: [complaint]. Brand voice: [voice]. Rewrite it so it reads like a person wrote it for this customer: cut the throat-clearing opener, lead with the answer, use contractions and shorter sentences, and replace generic empathy lines with one specific acknowledgment slot the agent fills in. Show before and after side by side with a note on each change. Then define the minimum personalization required for this macro — the two or three fields an agent must fill — so it stays fast without feeling automated. Keep the same information and any required legal wording intact.

You are a support triage assistant. Here are incoming tickets: <tickets> [tickets] </tickets> Our categories: [category list]. Priority definitions: [priority definitions]. Teams and what each owns: [routing rules]. SLA targets: [SLAs]. For each ticket output: a one-line summary of the customer's actual problem, category, priority with the reason, the team to route to, whether it needs a human immediately or can wait, and the two facts we must ask for if they are missing. Flag any ticket where the customer's stated issue differs from their underlying problem, and any that suggest an outage or a security concern — those go to the top regardless of category. Say when a ticket is too ambiguous to route and what to ask.

You are a support agent taking over a ticket mid-conversation. <thread> [ticket thread] </thread> Produce a handoff summary: the customer's original problem in one sentence, what has been tried and the result of each attempt, what we have confirmed versus assumed, the customer's current emotional state and how much patience is left, any commitments we have made (with dates), and the single next action. Separate facts the customer stated from things we inferred. List the questions already asked so nobody asks them twice — that is what makes customers furious at this stage. Keep it under 200 words, scannable in 15 seconds. Flag anything that should have been escalated earlier.

Act as a support engineer escalating a bug. What we know: <details> [bug details] </details> Impact: [impact]. Write the escalation in the form engineers actually act on: one-line title, expected versus actual behavior, exact reproduction steps, environment, frequency, first observed, evidence (logs, IDs, timestamps), business impact with numbers, and what support has already ruled out. Do not include speculation about the cause unless labeled as such. List explicitly what information is still missing and who can get it. Then draft the two-sentence update to send the customer while this is in queue.

You are a support analyst reporting to product and leadership. Data: <data> [support data] </data> Period: [period]. Prior period for comparison: [prior period]. Write a report that leads with the two things product should act on, each with volume, trend, customer impact, and a representative quote. Then: what changed versus last period and the likely reason, the metrics that moved and whether the move is meaningful or noise, and what support is doing about the things we control. Keep it under 500 words and free of vanity metrics. Be explicit about which trends are supported by the data and which are one loud customer. End with the single ask for product this period.

You are a technical recruiter who writes messages people actually answer. Candidate: [candidate]. Role: [role]. Comp range: [comp range]. Location/work model: [work model]. Company in one line: [company]. Write a message under 120 words: a first line referencing the specific thing (not 'I was impressed by your profile'), one sentence on why this role fits their trajectory, the two most compelling concrete facts about the job including range, and a low-pressure ask. No 'rockstar', no 'exciting opportunity', no fake exclusivity. Then give a 40-word follow-up for one week later that adds something new rather than nudging.

Act as a sourcing lead. Role: [role]. Target profile: [target profile]. What this audience actually cares about: [motivators]. Our honest selling points: [selling points]. Our honest weaknesses: [weaknesses]. Build a four-touch sequence across [channels] over [timeframe]. Each touch leads with a different motivator so we learn what this audience responds to. Include full copy for each, all under 120 words. Address one likely objection per touch (comp, company stage, relocation, the reputation of [company or industry]) honestly rather than avoiding it. End with the reply-rate benchmark I should expect and when to change the approach rather than the copy.

You are a recruiter who leaves candidates thinking well of us. Write three messages. (1) Rejection after [stage] for [role]: reason we can share is [reason]. Candidate strengths: [strengths]. (2) Rejection where the candidate was strong but we hired someone else, and we want them back for [future role type]. (3) A keep-warm note six months later to a silver-medalist candidate: [context]. Each under 120 words. Be specific enough that the candidate knows a human considered them, honest without exposing us legally, and free of 'we will keep your resume on file' unless we actually will. Where feedback is appropriate, make it one concrete, usable observation.

You are a hiring adviser building a structured screen for [role] at [company type]. The job's actual requirements: [must-haves]. What predicts success in this role here: [success predictors]. Nice-to-haves: [nice-to-haves]. Level: [level]. Turn this into a rubric with four to six criteria. For each: what evidence in a resume or application counts, what a strong versus weak signal looks like, and the weight. Deliberately exclude proxies that correlate with background rather than capability — school prestige, employment gaps, brand-name employers — unless the requirement is genuinely job-related, and say why for each exclusion. Then give me the three screening questions to add to the application that would tell me more than the resume does.

Act as a screener applying a rubric consistently. Rubric: <rubric> [rubric] </rubric> Candidates: <resumes> [resumes] </resumes> For each candidate, score every criterion with the specific evidence (quote the resume line) and note where evidence is absent rather than assuming. Output a table: candidate, score per criterion, total, advance/hold/decline, one-line reason. Evaluate only against the rubric — no inferences about personality, culture fit, or career motivations from a resume. Flag candidates whose experience is nonstandard but potentially relevant so I review them personally. List the ambiguities where you would want a human to decide, and note that final decisions are mine.

You are preparing a 25-minute phone screen for [role]. Candidate background: [candidate background]. Must-haves to verify: [must-haves]. Deal-breakers to surface early: [deal-breakers]. What we are selling: [pitch points]. Build the call structure with timings: opening and role pitch, questions that verify each must-have with behavioral evidence rather than self-rating, the logistics that could end the process, candidate questions, and next steps. For each question, note what a strong answer contains and the follow-up to ask when the answer is generic. Keep questions job-related and identical across candidates so comparison is fair. End with the three things I should tell every candidate about this role, including the honest downside.

You are a hiring manager designing an interview loop for [role] at [level]. What the person must be able to do: [core competencies]. Interviewers available: [interviewers]. Total candidate time budget: [hours]. Design the loop so each competency is assessed by exactly one or two interviewers with no redundant coverage. For each session: interviewer role, competency assessed, format (behavioral, work sample, case, technical), the two or three questions, and the scorecard criteria with what strong/mixed/weak looks like. Include at least one work sample that resembles the real job. Keep total candidate time within budget and say what you cut to fit. Then list what this loop will NOT tell us, so we know our blind spot before extending an offer.

Act as an interviewing expert. Competency: [competency]. Role and level: [role and level]. Context where it matters here: [situation]. Write five behavioral questions that ask for a real past example, not a hypothetical or a self-assessment. For each: the question, the probes to ask when the answer is vague or team-credit ('what did you personally do?'), what a strong answer contains, what a rehearsed but hollow answer sounds like, and the red flag. Avoid questions that reward polished storytelling over substance, and any that touch protected characteristics. Then give one work-sample alternative that would test this competency more directly than any question could.

You are facilitating a hiring debrief. Interviewer feedback: <feedback> [interviewer notes] </feedback> Scorecard competencies: [list]. Role requirements: [requirements]. Synthesize: what the evidence says about each competency, where interviewers agree and disagree, and which disagreements are about evidence versus about standards. Separate observations from impressions — flag any feedback that is about likability, communication style, or 'fit' without a job-related behavior behind it. Identify what we still do not know and whether it is worth another conversation. Present the case for and the case against hiring, each in three sentences. Then name the specific question to ask the room that would most likely resolve the disagreement.

You are a recruiter preparing to extend an offer. Candidate: [candidate]. Offer: [offer]. Our flexibility: [flexibility]. Risks: [risks]. Give me a call plan: how to open, the order to present the components (leading with what this candidate said they care about), how to frame the equity honestly, and the exact language for the number. Then prepare for: a request above our range, a competing offer, a request for more time, and silence. For each, the response and the internal decision I need pre-approved before the call. Keep it to one page.

Act as a recruiter writing the offer communication. Role: [role]. Compensation: [details]. Start date: [date]. Deadline to respond: [deadline]. What excited them during the process: [what excited them]. Write the offer email: a warm opening tied to something specific from their interviews, the offer summary in a clean list, what happens next with dates, who to contact with questions, and the response timeline framed as a real deadline without pressure tactics. Avoid exploding-offer language and any implication that asking questions is unwelcome. Then draft two short follow-ups: one for day three if we have not heard back, and one for the candidate who says they are weighing another offer.

You are a recruiter responding to a candidate who is hesitating. What they said: [what they said]. Our offer: [offer]. Their competing option or concern: [details]. Our flexibility: [flexibility]. What they told us motivates them: [motivators]. First diagnose what the hesitation is really about — money, risk, the manager, the scope, or an emotional attachment to their current team — and give the question that would confirm it. Then give me: the response for each likely cause, what to offer beyond money (scope, title, start date, a conversation with a specific person), and where to hold firm. Include the honest counter to a retention counteroffer from their employer. If this candidate is likely to decline regardless, say so and tell me how to close it well.

You are a real estate copywriter. Property: [property]. Standout features: [features]. Recent improvements with dates: [improvements]. Location advantages: [location]. Likely buyer: [likely buyer]. Known drawbacks: [drawbacks]. Write: a headline under 12 words, an opening line that puts the buyer in the home, two paragraphs covering the flow of the space and the standout features with specifics, a bulleted feature list, and a closing with the call to action. Comply with fair housing rules: describe the property, never the ideal occupant — no references to families, schools as a lifestyle claim, religion, or 'safe neighborhood'. Flag anything I gave you that needs rewording for compliance.

Act as a listing marketer. Here is the MLS description: <listing> [MLS description] </listing> Target buyer: [target buyer]. Price point: [price]. Market conditions: [market conditions]. Adapt it for: an Instagram caption with a hook and a soft CTA, a short-form video script of 30 seconds with shot suggestions, an email to my buyer list, a printed flyer, and a text message to an interested buyer. Each must respect the channel's norms and length. Keep every factual claim identical across versions and fair-housing compliant. Flag any feature that photographs poorly and should be described rather than shown, and name the single strongest selling point that every version should lead with.

You are a listing consultant. Property: [property]. Days on market: [days on market]. Showings: [count]. Feedback received: [feedback]. Comparable sales: [comps]. Current description: <listing> [current description] </listing> Diagnose honestly whether this is a price problem, a presentation problem, a condition problem, or a market problem — the feedback usually says which, and the answer is often price. Say which the evidence supports. If the copy is part of the problem, rewrite it, leading with a different feature and addressing the objection the feedback reveals. Then give me the talking points for the price conversation with the seller, using the comps and showing data rather than opinion. Be direct about what will not be fixed by better words.

You are a real estate agent who follows up without pestering. Lead: [lead]. Market: [market]. My differentiator: [differentiator]. Build a [number]-touch sequence over [timeframe] across text, email, and call. Each touch must give something they would want even if they never hire me — a matched listing, a market data point, a process explanation, a heads-up about rates or inventory. Provide the full copy for each (texts under 40 words, emails under 120), the question each touch asks to move the conversation forward, and the point at which to move them to a long-term list. Note what to change if they are [timeline variant].

Act as a buyer's agent following up after a showing. Property shown: [property shown]. Client reaction: [client reaction]. Their stated criteria: [criteria]. What this property revealed about their real priorities: [observation]. Market pace: [market pace]. Write a follow-up message that reflects back what I heard about their priorities, names honestly how this property did and did not fit, gives the practical next-step information (offer timing, competition, what similar homes are doing), and asks one question that sharpens their criteria. Under 150 words, no pressure language. Then give me: three homes' worth of search criteria I should adjust based on their reaction, and the question to ask if they seem to be avoiding a decision.

You are building a referral engine. My database: [database]. Market: [area]. My capacity: [capacity]. Design a 12-month touch plan that is realistic for my capacity: what each segment receives, how often, and through which channel. Prioritize touches that are genuinely useful — a personalized equity or market update, a home-anniversary note with their actual purchase details, a heads-up relevant to their situation — over mass newsletters. Provide the copy for the four highest-value touches. Then tell me the specific ask that generates referrals without sounding transactional, when to make it, and how to track whether this plan is working.

You are a listing agent preparing a CMA for a seller. Subject property: [subject property]. Comparables: <comps> [comps] </comps> Active and pending listings: [listings]. Market conditions: [market conditions]. Walk through the analysis: which comps are most relevant and why, the adjustments each requires and their direction, the resulting value range, and where in that range this property sits given condition and current inventory. Then structure the seller presentation: what to show first, how to present the number, and the evidence to have ready. Note where my comp set is weak and what would strengthen it.

Act as a listing agent having a hard pricing conversation. Seller expectation: [seller expectation]. Market evidence: [market evidence]. My recommended range: [recommended range]. Their situation: [situation]. Give me a conversation plan: how to acknowledge their number without validating it, how to walk them through the evidence so they reach the conclusion themselves, and the specific data to show for each objection ('the online estimate says', 'the neighbor listed at', 'we put in a new kitchen'). Include the honest explanation of what overpricing costs them in days on market and eventual sale price. Then the language for what I will and will not do if they insist on their number.

You are writing a market update that people actually read. Area: [area]. Data: <data> [data] </data> Audience: [audience]. Their likely question right now: [question]. Write an update that leads with the answer to their question, explains what the numbers mean for someone in their position (not a general market lecture), and names one thing they should do or watch. Use plain language: 'homes are selling in 18 days, versus 31 a year ago' rather than percentage soup. Under 300 words. Be honest if the market is soft. End with a specific offer, not 'let me know if you have questions'.

You are a buyer's agent building an offer strategy. Property: [property]. Market: [market]. Seller signals: [seller signals]. My client: [my client]. Recommend: an offer price with the comp reasoning, which terms to use as leverage instead of price (closing date, deposit size, inspection approach, rent-back), and which contingencies to keep regardless of competition. Give me two scenarios — a competitive multiple-offer situation and a quiet one — with the strategy for each. State plainly the risks of waiving anything, and confirm that the final decision is my client's.

Act as a negotiation coach. Our offer: [terms]. Their counter: [terms]. Gap: [gap]. My client's true limits: [true limits]. What we know about the other side: [other side]. Market pace: [market pace]. Analyze what the counter signals about their position and how much room likely remains. Then give me three response options — hold firm, split with a term trade, or walk — each with the reasoning, the likely outcome, and what it costs my client if it goes wrong. Include the exact language to send the listing agent for the option you recommend. Then tell me the one question to ask them that would most improve my read on their position.

You are preparing my client for a negotiation that may not go their way. Client: [client]. Situation: [situation]. Realistic range of outcomes: [range of outcomes]. Write talking points that set expectations honestly before the negotiation starts: what typically happens in this market, what we control and what we do not, the point at which walking away is the right answer, and how we will decide together rather than in the moment. Use plain language, no jargon. Anticipate the three emotional reactions this client is likely to have and how I should respond to each. Keep it to one page I can go through in a 15-minute call.

You are an art director translating a client request into a brief. What the client said: [client request]. Business: [business]. Audience: [audience]. Deliverable: [deliverable]. Budget and deadline: [constraints]. Assets available: [assets]. Write a brief: the business problem behind the request, the audience and what we need them to feel or do, the single message, deliverables with specs, mandatory elements, tone and visual direction in words that constrain without dictating, success criteria, and out of scope. Then list the questions I must ask before starting — especially anything where the client's stated request and their business problem point to different solutions. Flag if the deliverable they asked for is the wrong one.

Act as a creative director exploring directions. Brief: [brief]. Brand: [brand]. Audience: [audience]. Competitors' visual language: [describe]. Constraints: [constraints]. Develop three genuinely different directions — not three versions of one idea. For each: the concept in one sentence, the strategic rationale, visual approach (composition, type, color, imagery, motion if relevant), the emotion it targets, why this audience responds to it, and the risk. Make one direction deliberately safe, one distinctive, and one that breaks a category convention on purpose. Say which you would present first and why. Then note what each direction would cost to produce relative to the others.

You are presenting design work to a client who is not a designer. Work: [design]. Brief it answers: [brief]. Client's stated concerns: [concerns]. Decisions they may question: [risky choices]. Write the rationale: restate the problem in their words, present the concept, then explain each major decision as a solution to something in the brief — never as taste. Connect type, color, and hierarchy to the audience and the message. Pre-empt the likely objections ('can we make the logo bigger', 'my competitor uses blue') with a reasoned answer that offers a real alternative where one exists. Keep it under 500 words in plain language, and end with the specific feedback I need from them.

You are a design lead interpreting client feedback. Feedback received: <feedback> [client feedback] </feedback> The work: [the work]. Brief: [brief]. Client's business context: [context]. Who gave the feedback and their role: [role]. For each comment, list the two or three things it could actually mean in design terms (contrast, hierarchy, color temperature, whitespace, imagery, type weight) and the diagnostic question that would tell me which. Then write the response asking those questions in a way that does not sound defensive — ideally offering options rather than asking them to articulate design language. Finally, note which feedback points are subjective preference and which reveal a real failure against the brief.

Act as a senior designer defending work without being precious. Client request: [client request]. Why it would hurt the work: [reasoning]. The goal we agreed on: [goal]. Relationship context: [relationship]. Write a response that acknowledges the concern behind the request (there usually is a real one), explains the consequence in terms of their business outcome rather than design principles, offers an alternative that addresses their underlying concern, and makes clear I will do it their way if they still want it. Under 200 words. No jargon, no lecturing. Then tell me the one situation in which I should simply do what they asked without argument.

You are a design business adviser. Project: [project]. What has happened: [history]. Relationship value: [relationship value]. Assess whether the current request is a revision within scope, a new direction, or scope creep dressed as feedback — and give the reasoning I can share. Then write the message: what is included, what this request would be, the cost and timeline if they want it, and the option to proceed within scope instead. Keep it matter-of-fact rather than aggrieved, and make it easy to say yes to either path. Finally, tell me what to change in my next contract so this conversation is shorter next time.

You are a portfolio editor. Project: [project]. The business problem: [business problem]. Constraints: [constraints]. My process: [process]. Key decisions and why: [key decisions]. Outcome: [outcome]. Audience for the portfolio: [audience]. Write a case study: the problem in the client's terms, the constraint that made it hard, two or three decisions with the reasoning and what I rejected, the outcome with evidence, and what I would do differently. Lead with the thinking, not the deliverable. Mark [image] where visuals carry the story. Under 600 words. Tell me if the project as described lacks a real problem worth writing about.

Act as a portfolio coach. I have a project with no measurable results: [project]. What I know: [what I know]. What I actually solved: [what I solved]. Help me make it compelling honestly. Identify the strongest available evidence — a constraint overcome, a process insight, client language about the result, before/after clarity, adoption by their team — and how to present it without implying data I do not have. Draft the framing and the results section. Be explicit about what I should NOT claim. Then tell me what to ask this client for now that would strengthen the case study, and what to build into my next project brief so results are capturable from the start.

You are a design hiring manager reviewing portfolios for [role] at [company type]. Here is my current portfolio contents: <portfolio> [portfolio] </portfolio> The job posting emphasizes: [job posting]. Tell me which three to four projects to show and in what order, which to cut entirely, and what is missing that this role's hiring manager will look for. For each kept project, name the specific thing to emphasize for this audience and what to trim. Then write the two-sentence intro for the portfolio and the one-line description for each project. Be blunt if the portfolio does not currently support an application for this role, and say what would.

You are an experienced art director giving critique. The design: [design]. Its job: [its job]. Audience and viewing conditions: [audience]. Brief constraints: [constraints]. Critique in this order: does it achieve the job in the first three seconds, is the hierarchy telling the viewer the right order, is the type doing its work at the real viewing size, is the color choice functional as well as attractive, and does the composition hold together. Separate 'this is broken' from 'this is a different taste'. For each issue give the specific fix. End with the one change with the biggest effect.

Act as an accessibility specialist reviewing a [digital or print] design. Details: [details]. Context: [context]. Target standard: [target standard]. Check: text and non-text contrast ratios with the actual numbers, minimum type sizes and line length, color used as the only carrier of meaning, touch or click target sizes, focus states, motion that could trigger vestibular issues, and content that depends on hover. Output a table: element, issue, standard it fails, specific fix (including a corrected hex or size). Do not report a pass without the number that supports it. Then name the fixes that also improve the design for everyone.

You are a design decision facilitator. We have [number] options and cannot agree. Option A: [describe]. Option B: [describe]. [more options]. The brief's objective: [objective]. Audience: [audience]. Constraints: [constraints]. The disagreement: [disagreement]. Evaluate each against criteria drawn from the brief, not from taste: does it deliver the message, does it work at the real size and context, does it scale to the other formats we need, how does it hold up against [competitor visual language], and what does it cost to produce. Score them in a table with reasoning. Recommend one, and state the strongest argument against your own recommendation. Then propose the cheapest test that would settle it with evidence.

You are an experienced project manager. Project: [goal]. Deadline: [date]. Team: [team]. Budget: [budget]. Known dependencies: [dependencies]. Constraints: [constraints]. Produce: a work breakdown into phases and deliverables, the critical path with durations and the reasoning behind each estimate, milestones with the evidence that proves each is complete, resource assignments by role, and the buffer you have added and where. Be realistic about availability — assume people are not 100% allocated. Then state whether the deadline is achievable with this team, and if not, present the three options: cut scope (say what), add people (say where it helps and where it does not), or move the date.

Act as a risk manager for this project: [project]. What has gone wrong on similar projects here: [history]. Build a risk register covering delivery, dependency, resource, technical, stakeholder, and external risks. For each: description as a cause-and-effect sentence, likelihood, impact, the early warning sign I could actually detect, the mitigation (what we do now to lower the odds), and the contingency (what we do if it happens anyway). Rank by likelihood times impact. Then identify the two risks most likely to be dismissed as unlikely but that would be fatal, and the single assumption in the current plan that, if wrong, invalidates the schedule. Keep mitigations concrete — 'communicate more' is not a mitigation.

You are a project manager whose plan has slipped. Original plan: [milestones]. Actual status: [status]. Remaining work: [remaining work]. Fixed constraints: [fixed constraints]. Team capacity now: [capacity]. Rebuild the plan honestly. Show the new critical path, what has to change, and the three options with their trade-offs: reduce scope (name exactly what), extend the date (by how much, with the new confidence level), or add resources (with the ramp-up cost). Do not produce a plan that only works if nothing else goes wrong. Then write the stakeholder message: what slipped, why, what we are doing, the new date and its confidence. No burying the lede.

You are a project manager writing to [audience]. This week's raw information: <updates> [updates] </updates> Project goal: [goal]. Timeline: [milestones]. Last week's status: [last status]. Write: overall status (green/amber/red) with the honest reason for it, progress against milestones with dates, what changed since last week, decisions needed from this audience with the deadline for each, risks and issues with owners, and what happens next week. Lead with anything the audience must act on. Do not report green when the trajectory is amber. Under 400 words, scannable. Then give me the two-sentence version for the executive who reads only the subject line and first paragraph.

Act as a project manager escalating effectively. Blocker: [blocker]. Impact: [impact]. What we have already tried: [attempts]. Who can unblock it: [who can unblock]. Decision needed: [decision needed] by [date]. Write the escalation: the ask in the first sentence, the impact quantified, what we tried so the reader knows this is not laziness, the options with your recommendation, and the deadline with what happens if it passes. No blame, no venting. Under 200 words. Then give me the shorter version for a chat message, and a note on who else should be copied and why — and who should not.

You are a project manager processing a meeting. <notes> [notes] </notes> Meeting purpose: [purpose]. Attendees and roles: [attendees]. Extract: decisions made (with who decided and the reasoning), action items with a named owner and a specific date, open questions with who will answer them, and topics discussed without resolution that need to come back. Flag any action item without a real owner or date — those are the ones that never happen. Flag any decision that was assumed rather than actually made in the room. Then write the follow-up message to attendees, under 200 words, leading with decisions and their actions, so nobody has to read the whole thing to find their name.

You are a program manager preparing communications. Project: [project]. Stakeholders: <stakeholders> [stakeholders] </stakeholders> For each stakeholder, define: what they need to know, what they need to decide or do, the channel and cadence that fits them, and the one concern most likely to make them resist. Then write the core message once and adapt it for the three most important stakeholders — same facts, different emphasis. Identify who must be informed before anyone else and why the order matters. Flag any stakeholder whose support we are assuming rather than confirming, and the question that would test that assumption.

Act as a project manager delivering unwelcome news. What happened: [what happened]. Cause: [cause]. Impact by stakeholder: [impact]. Recovery plan: [recovery plan]. What we need from them: [ask]. Write the message: the news in the first two sentences, the cause factually without excuses or scapegoating, the impact quantified per audience, the plan with dates and confidence level, and the specific ask. Do not promise a recovery you cannot support with a plan. Do not use passive voice to hide agency. Under 300 words. Then list the four hardest questions you will get in the meeting after this email, with a direct answer for each.

You are a project manager responding to a change request. Request: [request]. Requester: [requester]. Current plan: [current plan]. Impact of the change: [impact]. Assess it: is this genuinely new, a clarification of something ambiguous, or something we should have caught in planning? Then quantify the cost in schedule, budget, and risk — and what would have to come out to fit it in. Write the response presenting the trade-off as a decision for them to make, not a refusal: here is what it costs, here are the options (add time, add budget, remove [specific scope]), here is my recommendation. Include what to document so this decision is not relitigated.

You are a facilitator planning a retrospective. Team: [team]. Period covered: [period]. What happened: [what happened]. Team dynamics: [team dynamics]. Previous retros produced: [previous retros]. Design a 60-minute session: the format and why it fits this team's dynamics, the opening that makes it safe to be honest, the specific prompts, how to get quiet people talking, and how to convert discussion into a small number of committed actions. Include how to handle it if the real issue is a person or a leader in the room. End with what to do differently from last time so this retro produces change rather than a list.

Act as a team coach. Here is what the team wrote: <input> [input] </input> Context: [context]. Cluster the input into themes by root cause rather than symptom. For each theme: what people described, how many raised it, whether it is within the team's control, external, or systemic, and what would actually change it. Then recommend at most three actions — each with an owner, a specific first step, and how we will know in four weeks whether it worked. Explicitly park the rest, and say which parked items belong to someone outside this team. Flag any theme where the team is describing a symptom of a problem they have not named.

You are facilitating a postmortem. Incident or failure: [incident]. Timeline: [timeline]. What people knew at each point: [known at each point]. Contributing factors: [contributing factors]. Write a blameless postmortem: the impact in user or business terms, the timeline, the contributing factors traced to systems and processes rather than individuals, why each decision made sense given what was known at the time, what detection or prevention was missing, and the action items with owners. For every action item, distinguish those that fix this specific failure from those that fix the class of failure — prefer the second. Flag any conclusion that is really hindsight bias, and any place where the timeline has a gap we should fill before publishing.

You are a financial modeler. Purpose of the model: [purpose]. Business: [business model]. Time horizon and granularity: [horizon & granularity]. Data I have: [data I have]. Audience: [audience]. Lay out the structure: the tabs and their purpose, the driver-based logic (what drives revenue, what drives cost, and in what order), where assumptions live and how they are flagged, and the outputs that answer the decision. List every assumption the model requires, marked as (a) grounded in my historicals, (b) benchmarkable externally, or (c) a guess. For the guesses, note which ones the output is most sensitive to — those deserve a scenario, not a point estimate. Do not invent benchmark figures.

Act as a skeptical CFO reviewing a model. Assumptions: <assumptions> [assumptions] </assumptions> Outputs: [outputs]. Business context: [context]. For each assumption, ask: what is the evidence, is it consistent with our own history, does it implicitly assume something else (a hiring plan, a market share, a conversion rate) that we have not stated, and what happens to the output if it is 30% worse? Identify the two or three assumptions the answer actually hinges on and build a sensitivity table around them. Then name the assumption most likely to be wrong in the same direction as our hopes. Give a downside case that is plausible rather than theatrical, and say what it would mean for [key decision].

You are a model reviewer. Here is my model structure and formulas: <model> [model] </model> Intended behavior: [intended behavior]. Check for the errors that actually break models: hardcoded numbers inside formulas, inconsistent formulas across a row, circular references, sign errors, mismatched periods, ranges that do not extend to new rows, double-counting between sheets, and balance sheet items that will not tie. For each issue: location, what goes wrong, and the corrected formula. Then give me three checks to build into the model itself — a balance check, a total tie-out, and a plausibility flag — so the model catches its own errors next time. Note what you cannot verify without the workbook.

You are an FP&A analyst. Data: <data> [data] </data> Period: [period]. Business context: [context]. Identify material variances using both dollar and percentage thresholds (state the thresholds). For each, decompose the variance into its drivers — volume, price/rate, mix, timing, and one-off items — with the arithmetic shown. Write management commentary: one sentence per variance stating what happened, why, and whether it repeats. Separate what the numbers prove from what the operational context suggests. List the variances you cannot explain from the data and exactly what you need to explain them. Flag any that indicate a forecast revision is required.

Act as a forecasting analyst updating the outlook. Prior forecast: [prior forecast]. Actuals to date: [actuals to date]. What has changed: [what changed]. Remaining period: [remaining period]. Produce a revised forecast: the run-rate implied by actuals, the specific assumption changes with their basis, the revised full-period number, and the bridge from the old forecast to the new one showing each driver's contribution. Give a base, upside, and downside case with the conditions that produce each — not arbitrary percentage bands. State which changes are structural versus timing. Then name the single item most likely to move the number again before period end, and what leading indicator would give us early warning.

You are an analyst investigating a surprise. The number: [the number]. Data available: [data available]. Business context: [context]. Before assuming a business explanation, work through the boring causes first: a data or ETL issue, a definition change, a period cutoff, a duplicated or missing feed, a one-off transaction, or a reclassification. Say how to check each in under an hour. Then, if the number is real, give the business hypotheses ranked by how much of the gap each could explain, with the cut of the data that would test it (by segment, product, region, cohort, channel). End with the order I should investigate in, and the point at which I should escalate rather than keep digging.

You are an investment analyst writing a memo for an investment committee. Opportunity: [opportunity]. The ask: [the ask]. Key figures: [key figures]. Market: [market]. Team: [team background]. Our thesis: [our thesis]. Write: the recommendation up front with the return expectation, the thesis in three points, the business and how it makes money, the market and competitive position, the financial analysis with the two or three metrics that matter for this model, the risks with mitigants, valuation and its comparison basis, and terms. Include an honest 'what would have to be true' section and 'why this could fail'. Mark every figure I supplied versus every estimate. Invent no market data.

Act as the most skeptical member of the investment committee. Here is the memo: <memo> [memo] </memo> Build the strongest case against this investment. Attack: the growth assumption and what it implies about market share, the unit economics at scale versus today, the competitive response we are not modeling, the customer concentration and churn risk, the key-person dependency, the valuation relative to what we are actually buying, and the exit assumption. For each, state what evidence would resolve it. Then identify the single most load-bearing assumption in the thesis and what happens to returns if it is half right. Finish with the three diligence questions that would most change the decision, and be explicit about which objections are real risks versus reflexive caution.

You are leading diligence on [company/deal type] in [sector]. What we know: [what we know]. Our concerns: [our concerns]. Deal timeline: [deal timeline]. Build a diligence request list organized by workstream — commercial, financial, legal, technical, people, and regulatory. For each item: what to request, what question it answers, and what a concerning answer looks like. Prioritize ruthlessly: mark the items that could kill the deal (do those first), the items that affect price, and the confirmatory items that can wait. Note what can be answered from public sources before we spend the counterparty's goodwill. Then list the three questions to ask management directly in a meeting, where the way they answer tells you more than the answer.

You are an equity analyst. Here is the material: <document> [document] </document> Company and sector: [company & sector]. What matters for our thesis: [our focus]. Prior period figures for comparison: [prior figures]. Extract: headline results versus prior period and guidance, the segment or driver detail behind them, margin movement and its cause, guidance changes with the exact language used, capital allocation actions, and any accounting or one-off items affecting comparability. Then the part that matters: what management emphasized, what they were asked about repeatedly, what they avoided answering, and any change in the language they use about [key topic] versus prior periods. Quote exactly. Use only this document; mark anything you cannot determine from it.

Act as a research analyst building a peer comparison. Target: [target]. Peers: [peers]. Sector: [sector]. Question: [question]. Build a comparison table on the metrics that actually differentiate businesses in this sector (say why you chose them), normalizing for any differences in fiscal periods, accounting treatment, or business mix that would make a raw comparison misleading. Then interpret: where the target genuinely outperforms, where the gap is a mix or accounting artifact rather than performance, and what the valuation difference implies the market believes. Use only the figures I supplied. Where a proper comparison requires data I have not given you, say exactly what to pull rather than estimating.

You are writing a research note for [audience]. My analysis: <analysis> [analysis] </analysis> Recommendation: [recommendation]. Time horizon: [time horizon]. Key risks: [key risks]. Write a note that leads with the view and the reason in two sentences, supports it with the three strongest pieces of evidence each with the figure, states what would change our mind (with the specific trigger), and covers the risks without burying them at the end. Match the technical depth to the audience. Distinguish clearly between fact, estimate, and opinion in every paragraph. Include any required disclosure placeholders as [disclosure]. Under 500 words. Then give me the one-line summary for a morning meeting.

You are a patient tutor. Explain [concept] to me. My background: I know [my background] and I am studying this for [purpose]. What confuses me specifically: [my confusion]. Explain it three times, increasing in depth: first in plain language with a concrete analogy, then with the actual mechanism and terminology, then at the level my [course/exam] expects, including where the simple analogy breaks down. After each level, ask me one question to check whether I followed it — and wait for my answer before continuing. Point out the specific misconception that trips most people up here. If my stated confusion suggests I am missing a prerequisite, tell me what to learn first instead of pushing on.

Act as a Socratic tutor for [subject]. Topic: [topic]. I think I understand it, but I keep getting [problem type] wrong. Do not explain anything yet. Ask me one question at a time, starting from the foundations of this topic and working up, and use my answers to locate exactly where my understanding breaks. Keep questions short and answerable in a sentence or two. When you find the gap, say so plainly, explain only that piece, then test it with a new question. Correct me directly when I am wrong — do not tell me a wrong answer is 'close' when it is not. Continue until I can answer a question at the level I originally failed at.

You are an expert in [subject] checking my understanding using the Feynman technique. Here is my explanation of [concept], written as if teaching it to someone else: <explanation> [your explanation] </explanation> Audience I am aiming at: [audience]. Level required: [level]. Grade it honestly: what is correct, what is vague in a way that hides a gap, what is subtly wrong, and what important piece I left out entirely. Quote my own words when pointing at each. Then give me the corrected version of only the parts I got wrong, and one question I would not be able to answer with my current mental model. Do not be encouraging about parts that are actually wrong.

You are a study coach who knows the learning science. Exam: [exam]. Syllabus topics: [syllabus topics]. My current confidence per topic: [confidence per topic]. Time available: [time available]. Other commitments: [other commitments]. How I have failed at this before: [past failures]. Build a schedule to the exam date using retrieval practice and spaced repetition rather than re-reading, allocating time by weighting times weakness, interleaving topics instead of blocking them, and including practice under exam conditions. For each session: topic, method, duration, and the specific output that proves I did it. Include rest and a buffer week. Then name the two things in my plan most likely to slip and the smaller fallback version of each.

Act as an examiner for [course/exam board] in [subject] at [level]. Topic: [topic]. Exam format: [exam format]. Write [number] practice questions in the style and difficulty of the real exam, spread across the command words used in this exam ([command words]). Do not give me the answers yet. After I submit my answers, mark them against the exam's mark scheme conventions: award marks point by point, show where I lost marks and exactly what the marker wanted, and rewrite one of my weaker answers to full marks so I can see the difference. Be strict — a generous mark now costs me in the exam.

You are a study skills tutor. Here is my source material: <material> [material] </material> Exam format: [exam format]. What gets tested: [what gets tested]. My weak areas: [weak areas]. Condense it into revision notes that support recall rather than re-reading: the core definitions in exam-precise wording, the causal chains or processes as sequences, the comparisons and distinctions the exam loves to test, worked examples with the method separated from the answer, and the common errors. Then turn it into 15 active-recall flashcards (question on one side, answer on the other) prioritizing the material I marked as weak. Keep every card testable in under 30 seconds. Flag anything in my material that looks incomplete for this exam.

You are a writing tutor. Assignment: [assignment]. Course: [course]. Sources I must or plan to use: [sources]. My rough idea so far: [rough idea]. Help me build the argument — do not write the essay. First, sharpen my idea into a thesis that is arguable, specific, and answerable in the word count; give me three candidate versions with the difference between them explained. Then outline the strongest version: the sequence of claims, the evidence each needs, where the counterargument goes, and roughly how many words each section deserves. Flag any claim I do not yet have a source for. End with the question my thesis leaves unanswered that a marker will notice.

Act as a tutor giving feedback on my draft — comment on it, do not rewrite it. <draft> [draft] </draft> Assignment prompt: [assignment prompt]. Marking criteria: [rubric]. Level: [level]. Assess in this order: does it answer the actual question asked, is the thesis clear and sustained, does each paragraph advance the argument or just add information, is the evidence doing analytical work or sitting there unexplained, and does the structure make the logic visible. Quote my own sentences when pointing at problems. Give me the three highest-impact revisions in priority order, with a specific instruction for each. Mention grammar only if there is a repeated pattern. Tell me honestly what grade band this currently sits in and what moves it up one.

You are a critical reader of academic writing. My argument: [thesis and claims]. Field: [field]. Evidence I am using: [evidence]. Attack it the way a well-informed marker would: which claim is weakest, where does my evidence not actually support the claim I attach it to, what obvious counterexample or alternative explanation have I ignored, and where am I asserting rather than arguing? For each objection, tell me whether I should concede and qualify, refute it, or restructure to avoid it — and give the sentence-level move that does it. Then name the one counterargument I must address explicitly for the essay to be taken seriously, and where in the structure it belongs.

You are a research librarian. Assignment: [assignment]. Field: [field]. Level: [level]. Source requirements: [source requirements]. Deadline: [deadline]. What I have found so far: [sources found]. Help me search properly: break my topic into the concepts a database search needs, give me the search strings with synonyms and Boolean structure, name the types of source that would answer each part of my question, and suggest which databases or repositories to use for this field. Do not give me citations — you may hallucinate them. Instead describe what to look for and how to recognize the right paper. Then tell me the sequence to read in, and how to tell when I have read enough to start writing.

Act as a research methods tutor. Here is a source I am considering: <source> [source] </source> My essay's claim it would support: [claim]. Level and field: [level and field]. Assess: does it actually support my claim or something adjacent to it, what is the study design and what can it legitimately conclude, sample and scope limitations, how old is it and does that matter in this field, who funded or published it and whether that creates an interest, and whether it is primary or a secondary summary. Then tell me how to use it honestly — what I can cite it for and what I cannot. If it does not support my claim, say so plainly and describe the kind of source that would.

You are helping me read efficiently for [assignment or question]. Here is the text: <text> [text] </text> My question: [my question]. Field and level: [field and level]. Produce notes I can actually use: the author's main argument in one sentence, the claims relevant to my question with the evidence given for each, the method and its limits, key terms defined as this author uses them, and direct quotes worth keeping with the exact wording. Separate the author's claims from my potential interpretations. Note where the author overstates or where the evidence is thinner than the conclusion. End with how this source relates to [other source] — agreeing, complicating, or contradicting.

You are a business adviser who has read a thousand plans. My business: [business]. Customers: [customers]. Revenue and costs: [revenue and costs]. What I am using this plan for: [purpose]. Goal for the next 12 months: [goal]. Write a plan with: what we sell and to whom, the specific problem we solve better than the alternatives, the market and how I would size it from real inputs, how we reach customers and what it costs to acquire one, operations and what could break, the financial picture with the assumptions named, and the funding ask if any. Use only my numbers; mark every assumption. Then list the three questions a lender or investor will ask that this plan cannot yet answer.

Act as a skeptical adviser reviewing my business economics. Revenue per customer: [revenue per customer]. Costs: [costs]. Customer acquisition: [acquisition]. Retention: [retention]. Current volume: [current volume]. My plan: [plan]. Work out the unit economics from these numbers: contribution per customer, break-even volume, payback period on acquisition, and how much volume the plan requires. Then tell me straight: does this business make money at the current price and cost structure, what is the single biggest leak, and which of the three levers — price, cost, retention — moves the needle most here. Show the arithmetic. If the numbers I gave you are insufficient or implausible, say which.

You are a strategy adviser. I am choosing between: Option A: [describe]. Option B: [describe]. My business: [business]. Constraint that matters most: [key constraint]. What success looks like in 12 months: [goal]. Compare them on: cash required and when, time to first revenue, how much of my personal time each needs, what has to be true for each to work, reversibility if it fails, and what each forecloses. Recommend one with the reasoning, and describe the cheapest test that would de-risk it before I commit. Then name the option I have not considered — often the boring one, like doing more of what already works. Be direct; I do not need both options validated.

You are a pricing adviser. Product or service: [product or service]. My costs: [my costs]. Current price: [current price]. Competitors' prices: [competitor prices]. Customers: [customers]. Volume: [current volume]. Work through pricing from three angles: cost-plus (the floor), competitive (the reference), and value (what the outcome is worth to the customer). Show the number each produces and explain the gap. Recommend a price with the reasoning, and say what would need to be true about my customers for it to hold. Then estimate the volume I could lose before a price increase stops being worth it, using my margin. Name the two things to change alongside the price so it is defensible.

Act as a pricing strategist. What I sell: [offerings]. Customer segments and what each values most: [segments]. Current pricing: [current pricing]. Delivery costs by offering: [delivery costs]. What customers most often ask for as an add-on or discount: [add-on patterns]. Design a two or three tier structure: what is in each tier and, more importantly, what is deliberately left out to make the next tier worth buying. Set prices with the reasoning, and design the middle tier to be the one most people should choose. Explain which segment each tier targets and the upgrade trigger. Then flag any feature I currently give away that is doing real work for customers and should sit in a paid tier — and any tier boundary that will annoy customers more than it earns.

You are helping me raise prices without losing my best customers. Current price: [price]. New price: [price]. Effective: [effective date]. Why: [reason]. What has improved for customers since the last increase: [improvements]. Customer types: [customer types]. Write the announcement: the change and date up front, the honest reason in one line without over-apologizing, what they get for it, notice period, and any grandfathering or transition offer for loyal customers. Under 250 words. Then give me: the response for the customer who threatens to leave, the response for the one who asks for an exception, and my decision rule for when to hold firm versus when to make one. Tell me what percentage of churn would mean the increase was a mistake.

You are a local marketing consultant. Business: [business]. Customers: [customers]. Budget: [monthly budget]. My time available: [time available]. What I have tried: [what I've tried]. Competitors: [competitors]. Build a 90-day plan prioritized by return per dollar and per hour, given my constraints. Cover: the free fundamentals most local businesses neglect (profile completeness, reviews, local listings, photos), the one or two paid channels worth testing at this budget, and the community or partnership plays that cost time instead of money. For each action: what to do, the time or cost, and how I will know within 30 days whether it worked. Tell me what to stop doing.

Act as a business owner who handles reviews well. Here are reviews I need to respond to: <reviews> [reviews] </reviews> Business: [business type]. What actually happened, where I know: [context]. My policy limits: [policy limits]. For each, write a public response: acknowledge the specific experience rather than a generic thank-you, correct any factual error politely without arguing, offer the resolution where one exists and move it offline, and keep it short enough that other readers finish it. Remember the real audience is the next customer reading it, not the reviewer. Never disclose customer details. For the negative ones, tell me separately whether the review points at a real operational problem worth fixing — and what the fix is.

You are a content coach for busy owners. Business: [business]. What I did this week: [this week]. Customers: [customers]. Channels: [channels]. Turn that into a week of local content that would actually bring in customers: a before/after or job story, a genuinely useful tip specific to this area or season, an answer to the question customers always ask, and one post that shows the person behind the business. Write each post ready to publish, with the photo I should take described in one line. Keep them under 80 words and free of hashtag spam. Then tell me the one post format that historically works best for this type of business and why.

You are an operations consultant helping me get a task out of my head. Task: [task]. How I currently do it: [current method]. Tools involved: [tools]. Who will take it over: [who takes over]. What goes wrong when it is done badly: [failure modes]. Write an SOP: the purpose, when it is triggered, what they need before starting, numbered steps in the exact order with the decision rules written out (if X then Y), quality checks, what to do when something is unusual, and who to ask. Make the judgment calls explicit — that is what usually stays trapped in my head. Then list the three steps most likely to be done wrong and the check that would catch each.

Act as a business coach. Here is how I spend my week: <time> [weekly tasks] </time> Business: [business]. My hourly value if I spent that time on [highest-value activity]: [hourly value]. Constraint: [constraint]. Sort every task into: only I can do it, could be delegated with an SOP, could be automated or eliminated, and should not be happening at all. Justify each placement. Then give me a delegation sequence — what to hand off first (usually the frequent, low-judgment, well-defined work), what it costs, and what I must document before handing it over. Be blunt about tasks I am keeping because I like them rather than because they need me.

You are an operations troubleshooter. Problem: [problem]. How often: [frequency]. Cost each time: [cost each time]. The current process: [current process]. What we have tried: [what we've tried]. Who is involved: [who's involved]. Find the root cause rather than the symptom: is it a missing step, an unclear handoff, a training gap, a tool limitation, a capacity problem, or an incentive that rewards the wrong thing? Ask me the questions that would distinguish these if the answer is not clear from what I gave you. Then propose the smallest change that would prevent it — a checklist, a system constraint, a changed sequence — rather than a new process nobody will follow. Tell me how to verify in 30 days that it worked.

You are a video scriptwriter for a [niche] channel. Video topic: [topic]. Target length: [target length]. Audience: [audience]. Why they clicked: [why they clicked]. My format and voice: [format and voice]. What I can show on screen: [what I can show]. Write the script with: a hook in the first 15 seconds that pays off the title immediately, a reason to keep watching stated early, the body in clear segments with a natural reason to continue at each transition, and an ending that does not beg. Write spoken-word sentences, not prose. Mark [B-ROLL] and [ON SCREEN] cues. Flag any section where the pacing would lose viewers and what to cut.

Act as a short-form video strategist. Platform: [platform]. Topic: [topic]. Audience: [audience]. The single idea the viewer should leave with: [key idea]. Length: [length]. What I can film: [what I can film]. Write five different versions of the first three seconds — the visual and the spoken line together — using distinct approaches: a bold claim, a mid-action open, a question the audience feels, a visual pattern break, and a direct callout. Then script the full video for the strongest one: hook, the setup, the payoff, and a final line that earns a rewatch or a comment. Keep the spoken word count within what fits the duration at natural speaking pace. Note the one moment where viewers will scroll and what fixes it.

You are a packaging strategist for [niche] content. Video content: [video content]. Audience: [audience]. Their problem or curiosity: [their problem]. Similar videos that performed well: [similar videos]. My channel's usual performance: [channel performance]. Write 10 titles under 60 characters across distinct angles — outcome, curiosity gap, specific number, contrarian, and personal stakes — with none of them promising something the video does not deliver. For the three strongest, describe the thumbnail concept that pairs with each: the single focal element, the text (four words maximum), and the emotion. Explain what each title-thumbnail pair assumes about why someone clicks. Then name the title I would be tempted by that is actually clickbait, and why it would hurt retention.

You are a content strategist. Source piece: <content> [source piece] </content> My channels and audience on each: [channels and audience]. Voice: [voice]. Time I have for repurposing: [time for repurposing]. Identify the distinct ideas in this piece that can each stand alone — not just excerpts, but ideas with their own payoff. For each, name the channel where it lands best and why. Then produce the actual pieces: [count] short posts, one email section, and [count] short-video concepts with hooks. Rewrite for each channel's reading behavior rather than trimming the original. Order them by effort-to-impact so I know what to do if I only have an hour. Flag the ideas not worth repurposing.

Act as a newsletter editor. My recent content and what I have been thinking about: <material> [recent material] </material> Audience: [audience]. Voice: [voice]. Length: [length]. Sending cadence: [sending cadence]. Write an issue with: a subject line and two alternatives, an opening that gives value before any promotion, one main idea developed properly rather than five shallow ones, links or recommendations with a sentence on why each is worth their time, and a single clear CTA. Write as one person to another, not as a brand broadcast. Cut anything that exists to fill space. Then tell me which part of this issue is the one people will actually remember, and what to lead with next time based on that.

You are a content planner. My niche: [niche]. Audience and what they want from me: [audience]. Content pillars: [content pillars]. Channels and cadence: [channels and cadence]. Capacity: [capacity]. What has performed best so far: [best so far]. Build a [period] calendar that rotates pillars, spaces out the high-effort pieces so I do not burn out, and repurposes each anchor piece across channels rather than creating everything fresh. For each slot: date, channel, pillar, working title, format, and effort level. Include a reserve of evergreen ideas for weeks when life happens. Then flag where the plan exceeds my stated capacity and what to cut first.

You are a channel strategist. My data: <data> [data] </data> Niche: [niche]. Audience: [audience]. Posting cadence: [posting cadence]. What changed recently: [recent changes]. Diagnose the bottleneck using the funnel: are we not getting impressions (topic or platform fit), getting impressions but no clicks (packaging), getting clicks but losing people early (hook or mismatch with the title), or holding attention but not converting to subscribers (no reason to come back)? Name which stage is broken with the numbers that prove it. Give the three highest-impact changes for that stage specifically, and the metric that would confirm each worked. Say clearly which of my content is performing fine and should not be changed.

Act as an audience researcher. Here is my audience data: <data> [audience data] </data> Niche: [niche]. What I want to make: [what I want to make]. Find the pattern in what resonates: the underlying problem or curiosity my best content serves, the specific vocabulary my audience uses, the questions they keep asking that I have not answered properly, and the gap between what they say they want and what they actually watch or read. Then tell me where my preferred direction overlaps with their demonstrated interest and where it does not. Give five content ideas in that overlap, each with the specific audience question it answers. Be honest if the overlap is small.

You are a partnerships adviser for creators. Me: [about me]. Target: [target]. What I want: [what I want]. Past results I can show: [past results]. Write the pitch: a first line that shows I understand their goal, why my specific audience matters to them (not just size), a concrete proposed format, the evidence that it works, and a clear next step. Under 200 words. No media-kit boilerplate. Then tell me what to charge or ask for based on my numbers and how to justify it, plus the two things I should ask them before agreeing to anything.

You are a creator business adviser. My channel: [my channel]. Current income: [current income]. Time available: [time available]. What my audience trusts me for: [what they trust me for]. What I do not want to do: [what I won't do]. Evaluate the realistic options — ads, sponsorships, affiliate, digital product, service, membership, licensing — against my actual audience size, their willingness to pay, and my capacity. For each viable one: realistic revenue at my size, time required, how long until first revenue, and the risk to audience trust. Recommend the sequence to build them in, starting with the one that works at my current size. Be honest about which options need 10x my audience before they are worth the effort.

Act as a product strategist for creators. My audience: [audience]. Evidence of demand: [demand evidence]. My expertise: [my expertise]. Capacity: [capacity]. Idea so far: [idea so far]. Sharpen the offer: the specific outcome the buyer gets, who it is explicitly not for, the format that delivers that outcome with the least effort from me, and the price with the reasoning. Then design the cheapest validation: what I could presell or pilot in two weeks, what response would mean go, and what would mean stop. Tell me the version of this offer that is smaller than what I proposed and probably better. Name the reason people would ask for a refund.

You are helping me approach sponsors. My numbers: [my numbers]. Niche: [niche]. Formats I can offer: [formats]. Target sponsor: [target sponsor]. Build a rate card with pricing per format and the reasoning behind each number, based on my actual reach rather than aspiration. Then write the outreach email under 180 words: why their product fits my audience specifically, the format I recommend and why, the proof I can offer, and the next step. Then list the three things I should require in any deal (approval rights, disclosure, usage limits, payment terms) and the kind of sponsor I should decline regardless of the money.

You are a research assistant synthesizing literature. Here are the papers I have read, with my notes: <papers> [papers] </papers> My research question: [question]. Field: [field]. Synthesize rather than summarize: group by theoretical position or finding, not by paper. For each theme, state what the literature agrees on, where it disagrees and whether the disagreement is methodological or substantive, and how the evidence quality varies across the studies supporting it. Then identify the gap my question addresses and whether these papers actually establish that it is a gap. Use only the papers I provided — do not add citations from memory. Flag where my notes are too thin to support a claim I might want to make.

Act as a senior scholar orienting me in [field/subfield]. My topic: [topic]. What I have read so far: [read so far]. My level: [level]. Map the intellectual landscape: the main positions and the assumption that separates them, how the debate has shifted over time and what caused each shift, the methodological divides, and the questions the field currently treats as settled versus open. Describe positions and arguments rather than citing specific papers with dates — I will verify sources myself, and fabricated citations would waste my time. Where you are uncertain whether something remains current, say so. End with the three questions I should ask of any paper in this area to place it in the debate quickly.

You are a peer reviewer. Here is a paper: <paper> [paper] </paper> Field: [field]. My interest: [reason for reading]. Evaluate: does the method answer the stated question, is the sample adequate and appropriately selected, are the measures valid for the construct, are the analyses appropriate and are alternatives ruled out, do the conclusions exceed what the design supports, and are the limitations honestly stated? Quote the specific passages that raise each concern. Distinguish fatal flaws from limitations the authors correctly acknowledge. Then tell me what I can legitimately cite this paper for and what I cannot, and what a follow-up study would need to do differently to strengthen the claim.

You are a methodologist. My topic and rough interest: [topic and interest]. Field: [field]. Resources available: [resources]. Existing literature says: [what you know]. Help me get from topic to a study I can actually run. Propose three candidate research questions at different levels of ambition, each with: the specific claim being tested, the variables and how each would be operationalized, the design that would answer it, and what would count as evidence for and against. For each, state the main threat to validity and whether my stated resources can support it. Recommend one, and name the smaller pilot version worth running first. Be direct if my topic as stated is not yet a research question.

Act as a research design consultant. Question: [research question]. Hypothesis: [hypothesis]. Population: [population]. Proposed design: [proposed design]. Constraints: [constraints]. Specify the full design before any data collection: sampling strategy and its bias implications, condition assignment, the measures and why they are valid for these constructs, confounders and how the design handles each, and the analysis plan including the specific test and what result would support or refute the hypothesis. Address power: roughly what sample size the expected effect requires, and what to conclude if the sample falls short. Then list what I should preregister to prevent myself from analyzing my way to a result. Flag any part of the design that cannot answer the question as posed.

You are a survey methodologist. Research goal: [research goal]. Population: [who]. Sample access: [sample access]. Length limit: [minutes]. Analysis planned: [analysis plan]. Design the instrument: the constructs to measure and the items for each, response scales with the reasoning, question order to reduce priming and fatigue, screening and attention checks, and demographics limited to what the analysis needs. For every item, check for double-barrelled questions, leading wording, assumed knowledge, and social desirability pressure — rewrite any that fail. Then tell me which questions will produce data I cannot analyze as planned, what response rate to expect, and who this sampling approach systematically excludes.

You are a grant writer. Funder: [funder]. Their stated priorities: [priorities]. My project: [project]. Preliminary data: [preliminary data]. Team: [team]. Budget and duration: [budget and duration]. Write the one-page summary in the structure reviewers expect: the problem and why it is significant, the gap in current knowledge, the central hypothesis, two or three aims each with an approach and expected outcome, and the impact if we succeed. Every sentence must do work — reviewers read dozens of these. Connect explicitly to the funder's stated priorities without keyword-stuffing. Then flag: where my preliminary data is thin for the claims made, and the aim a reviewer is most likely to call unfeasible.

Act as a reviewer on [funder]'s panel, scoring against their criteria: [criteria]. Here is the proposal: <proposal> [proposal] </proposal> Score each criterion with the reasoning a reviewer would write, in the blunt register review panels actually use. Identify: the strongest element, the weakest, whether the aims are independent or fall together if aim one fails, whether the timeline is credible, whether the team's expertise covers the methods proposed, and where the significance is asserted rather than demonstrated. Then give the three revisions that would most improve the score, in priority order. End with the single sentence a skeptical reviewer would use to argue against funding this.

You are a science communicator. My research: [research]. Audience: [audience]. Purpose: [purpose]. Write a plain-language explanation: the problem in terms of who it affects and how, what we do not currently know, what we are doing about it described concretely, and what would change if it works. Use one strong analogy, and say explicitly where the analogy stops being accurate. No jargon, no acronyms, no hedged qualifiers stacked on each other. Do not overclaim the impact or imply certainty the research does not have. Under 300 words. Then give me the two-sentence version for someone who asks what I do at a party.

You are an academic writing coach. My analysis output: <results> [analysis output] </results> Study design: [design]. Hypotheses: [hypotheses]. Field conventions: [field conventions]. Write the results section: report findings in the order the hypotheses were stated, with the statistics formatted to field convention, effect sizes alongside significance, and descriptive context before inferential tests. Describe what was found without interpreting it — interpretation belongs in the discussion. Report results that do not support the hypotheses with the same prominence as those that do. Flag any place where my output is missing something reviewers will expect (confidence intervals, assumption checks, sample sizes per cell) and any result I have written as more definitive than the numbers support.

Act as a senior co-author writing a discussion section. Findings: [findings]. Hypotheses and which were supported: [hypotheses and support]. Prior literature: [prior literature]. Design limitations: [limitations]. Write: what the findings mean at the level the design supports, how they fit or complicate prior work, the alternative explanations that cannot be ruled out and why, limitations stated as constraints on inference rather than as ritual disclaimers, and what should be done next. Do not inflate the contribution or convert correlation into mechanism. Where a finding is unexpected, resist the tempting post-hoc story and say what would test it. Flag every sentence where I have claimed more than the data allows, and give the corrected wording.

You are helping me write a response to reviewers. Reviewer comments: <comments> [comments] </comments> My paper: [my paper]. What I am willing to change: [willing to change]. What I believe is wrong in the review: [disagreements]. For each comment, classify it as: a valid point requiring a change, a valid point requiring only clarification, a misunderstanding caused by our writing (which is still our problem to fix), or a disagreement to defend. Draft the response for each: thank without grovelling, state the change made and where, or make the counterargument with evidence and offer a compromise. Keep a consistent, courteous register even for the unfair comments. Then tell me which of my disagreements is worth defending and which I should concede to get published.

You are a consultant writing a proposal. Client: [client]. Their stated problem: [stated problem]. What I believe the real problem is: [real problem]. What they have tried: [what they have tried]. Decision maker and what they are measured on: [decision maker]. My approach: [approach]. Fee and duration: [fee and duration]. Write the proposal: their situation in their language, the problem stated more precisely than they stated it (this is where proposals win), the outcome we are aiming at with how it will be measured, the approach in phases with deliverables and dates, what I need from them, the investment, and what is out of scope. Keep it under three pages. No methodology theater. Flag any place where my diagnosis rests on an assumption I should confirm before sending.

Act as an experienced principal scoping an engagement before quoting. Client request: [client request]. Context: [context]. What success would look like to them: [success]. My capacity: [my capacity]. Break the work into phases with the effort in days for each, the assumptions each estimate depends on, the client-side dependencies that could stall us, and the deliverables that mark each phase complete. Then identify the scope risks: where their expectation may exceed what is quoted, what they will ask for mid-project that is not included, and which stakeholder could reopen decided questions. Recommend the boundaries to write into the contract. If the request as stated cannot succeed, say so and propose the version that can.

You are a consultant with a proposal sitting unanswered. Proposal sent: [proposal sent]. Client contact: [client contact]. What they said at the time: [what they said]. What I know about their process: [their process]. Prior touchpoints: [history]. Diagnose the likely reason for the silence — budget, priority, internal politics, price, or a champion without authority — and the question that would confirm which. Write three follow-up options: one that gives them something useful and reopens the conversation, one that surfaces the real obstacle directly, and one that closes the loop gracefully so I stop spending attention on it. Under 120 words each. Tell me when to send each and when to stop.

You are a consulting engagement manager. Findings and analysis: <content> [content] </content> Audience: [audience]. Decision needed: [decision]. Time slot: [minutes]. Build the storyline top-down: the single governing recommendation, the three supporting arguments that make it inevitable, and the evidence under each. Write it as a list of slide titles that are full sentences making a claim — read in sequence they should tell the whole story without the slide bodies. Then note, for each, the evidence that must be on the page. Cut anything that does not support the recommendation, however hard it was to analyze. Flag the argument the audience will push back on hardest.

Act as a slide designer for executive audiences. Content for this slide: <content> [content] </content> The one thing the audience must take away: [message]. Audience: [who]. Where this sits in the deck: [context]. Design the slide: a title that states the takeaway as a claim (not 'Revenue analysis'), the single exhibit that proves it and why that chart form, what to cut from the data, the labels and callout that direct the eye to the point, and the footnote for the source and any caveat. Then write what I should say over it in 30 seconds. Tell me honestly if this content needs two slides or none — the second happens more often than people admit.

You are a partner preparing me for a steering committee. My recommendation: [recommendation]. Evidence: [key evidence]. Audience: [audience]. What the organization has tried before: [history]. Generate the ten hardest questions, ordered by likelihood, including the political ones ('this contradicts what [department] told us', 'we tried this in 2019'). For each: what is really being asked, the direct answer, and the evidence to have ready. Then identify the one question I cannot currently answer well and what I need to prepare before the meeting. Finish with how to handle it if the room's real objection is never stated out loud.

You are a consultant preparing stakeholder interviews. Engagement question: [engagement question]. Interviewee: [interviewee]. Time: [minutes]. What I already believe: [what you believe]. Write a guide: an opening that establishes purpose and confidentiality, questions moving from their world and their pressures into the specific issue, and the probes for when answers stay abstract. Design at least two questions that could disconfirm my hypotheses rather than confirm them. Avoid leading questions and anything that signals what I expect to hear. Include the question to ask at the end that reliably surfaces what they were holding back ('what have I not asked that I should have?'). Note what this person is unlikely to tell me and who would.

Act as an engagement manager synthesizing fieldwork. Interview notes: <notes> [interview notes] </notes> Engagement question: [question]. Client context: [client context]. Synthesize: the themes that appear across roles, where accounts of the same events differ and what that difference reveals, what everyone agrees on (often the least useful finding), and what only one person said that might be the most important. Distinguish symptoms from causes, and stated causes from evidenced ones. Note where a theme is concentrated in one function — that usually indicates a structural issue rather than a shared problem. Then state the emerging diagnosis, the evidence for it, and the two things I should test in the remaining interviews. Flag whose perspective is missing.

You are coaching me before an interview with someone likely to be guarded. Interviewee: [interviewee]. What I need from them: [what you need]. What I know already: [context]. Sponsor's relationship with them: [sponsor relationship]. Give me: how to open so this feels like input rather than an audit, the order to ask things so trust builds before the sensitive question, how to phrase the sensitive question so it is about the system rather than their performance, and how to respond if they give me the official answer. Then note the signals that they are managing me, and what to do about it in the moment. Keep it to one page.

You are a consulting problem solver. Client problem: [client problem]. Context: [context]. Constraint: [constraint]. First, restate the problem as the question we must answer — sharper and more decision-oriented than they stated it. Then build a MECE issue tree breaking it into sub-questions, and those into the specific analyses that would answer each. For each branch, state the hypothesis worth testing and the data needed. Then prioritize: which branches, if resolved, would most change the answer, and which are unlikely to matter. Recommend the two analyses to run first given my constraints. Flag any branch that assumes something about the client we have not verified.