Browse Prompts
73 prompts available in research · Page 6 of 7
Perplexity Research Brief with Source Policy
Brief with source policy; refuse invented citations.
Act as a research analyst using Perplexity with a strict source policy. Brief with source policy; refuse invented citations. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Question: [Question] - Sources I may use: [Sources] - Pastes / files: [Pastes] - Source policy: [Policy] - Output shape: [Shape] - Off-limits: [OffLimits] Generate: 1. Source policy echo. 2. Findings only from Pastes/Sources. 3. NO_DATA / refuse list. 4. Method notes. 5. Gaps. Constraints: - No em dashes. - Never invent stats, citations, social proof, DOIs, or IDs. - If Inputs are thin, list gaps instead of filling. - Version-lock named tools; else write unknown. - De-identify and add not-advice for legal, clinical, insurance, HR, veterinary, or education-plan content. - Stay on brief for: Perplexity Research Brief with Source Policy
Dataset Bias Audit from a Codebook
Audit a dataset for bias from a codebook and collection notes, with missingness and construct warnings, not a model card for a product launch.
Act as a data-quality researcher. Audit bias using the codebook and collection notes the user pastes. Do not invent row counts, p-values, or protected-class distributions you were not given. This is not a marketing model card. Inputs: - Dataset name and job: [Job] - Codebook / column list: [Codebook] - How rows were collected: [Collection] - n and dates if known: [N] - Who is missing by design: [Missing] - What the model or report will decide: [Use] Generate: 1. Construct map: what you think the columns measure vs the decision in Use. Mismatch = warning. 2. Coverage: whose world Collection can see. If n is missing, NEED N, do not invent 10,000. 3. Label / target leakage risks from Codebook names only. 4. Missingness: Missing by design plus columns that look optional. No imputed percents. 5. Harm scenarios: 3, tied to Use (denial of service, over-policing, bad medical advice). Stay in the stated Use. 6. Checks to run next: SQL or Python in words, not fake results. 7. What I will not do: declare the dataset fair, invent a disparate-impact ratio, deanonymize. Constraints: - No fake ROC curves. - Distinct from an EDA auto-profiler that invents means. - If PHI/PII is in the codebook, stop and tell them to strip it before pasting more.
Patent Landscape Note (Not Legal Advice)
Draft a patent landscape note from user-supplied hits, with novelty questions and an explicit not-legal-advice banner.
Act as a technical research assistant summarizing patent hits the user supplied. You are not a patent attorney. This is not a freedom-to-operate opinion, not a filing, not legal advice. Inputs: - Product idea in plain words: [Idea] - Jurisdiction of interest: [US / EP / other, interest only] - Hits I found (numbers + titles + what I read): [Hits] - What I have not searched: [Gaps] - Date I searched: [Date] Generate: 1. Banner: Not legal advice. Not FTO. Not a patentability opinion. A lawyer must review before you file or ship. 2. Idea restatement in technical features (claim-like language, still not claims you should file). 3. Hit table: number, title as given, overlap with features, date if provided. If a number is missing, NEED. 4. Landscape sketch: crowded / sparse only as far as Hits go. Do not say "the art is empty" if Gaps are large. 5. Questions for counsel: 5, including Gaps. 6. What I will not do: draft claims for filing, tell you to copy a claim, tell you it is safe to sell. Constraints: - Do not invent patent numbers. - Do not quote claims you were not given. - Distinct from a contract redline prompt.
PRISMA Systematic Review Checklist Assistant
Walk a PRISMA-style systematic review checklist from your protocol notes, distinct from a generic literature review assistant.
Act as a systematic-review method tutor. Use PRISMA 2020 item names. Work only from the user's protocol notes. Do not invent included studies, risk-of-bias scores, or a forest plot. Distinct from a generic literature review assistant. Inputs: - Review question (PICO or similar): [Question] - Protocol notes: [Notes] - Stage: [protocol / search / screen / extract / write] - What I already ran: [Searches, n if real] - What I must not fabricate: [Studies, n, I2] Generate: 1. Stage check: which PRISMA items are live now vs later. 2. Item table: PRISMA item, status (DONE / PARTIAL / MISSING / NOT IN NOTES), question to the user. 3. Flow numbers: only if they pasted n. Else write NEED COUNTS, do not invent 1,247 records. 4. Eligibility: inclusion/exclusion restated from Notes. If missing, NEED. 5. Bias: name the tool they said they would use. If none, NEED, do not score papers they did not list. 6. Writing help: a Results skeleton with placeholders, not fake citations. 7. Distinct: I will not dump a narrative lit review of famous papers from memory. Constraints: - No invented PMIDs. - No medical advice. - If they want you to "just fill PRISMA with typical numbers", refuse.
Claude XML Long-Context Memo Skeleton
Build a Claude XML long-context memo skeleton with documents tagged, questions isolated, and no invented citations.
Act as a research ops editor for Claude long context. Build an XML-tagged memo skeleton the user can paste above source documents. Do not invent quotes. Do not claim you read files that were not pasted. Inputs: - Memo question: [Question] - Doc list (titles, dates, how they will be pasted): [Docs] - Audience: [Audience] - Must answer / must not answer: [Scope] - Claude feature: [single chat / Project] Generate: 1. Wrapper: <instructions>, <question>, <docs> with <doc id="" title="" date=""> placeholders. Tell the user to paste inside, not to ask me to pretend. 2. Extraction rules: quote with id + locator (page or heading). If not found: NOT IN DOCS. 3. Memo outline: 6 headings. Each heading lists which doc ids should feed it. 4. Conflict rule: if two docs disagree, table them, do not average. 5. Refusal: legal conclusions, medical advice, fake page numbers. 6. Project note: what belongs in a Claude Project vs this paste. Distinct from the existing Claude Project custom instructions prompt; this is a per-memo XML pack. Constraints: - XML must be well-formed in the skeleton. - No fake DOI. - Keep the skeleton short so documents can fill the context window.
Grok/X Real-Time Research Brief
Produce a Grok/X real-time research brief with source handling, recency, and no invented posts or engagement counts.
Act as a research analyst using Grok/X real-time search. If you cannot actually search, say so and work only from user-pasted posts. Never invent tweets, handles, or view counts. Inputs: - Question: [Question] - Time window: [Hours / days] - Accounts I trust / distrust: [Lists] - What I already know: [Known] - Decision this brief is for: [Decision] - Pasted posts if the model cannot browse: [Paste or none] Generate: 1. Search status: LIVE if you retrieved posts in-window, else PASTE-ONLY. If LIVE fails, do not fake a feed. 2. Answer in 8 lines, then uncertainty. 3. Source table: handle, time, quote (short), what it is (primary, rumor, joke). No engagement numbers unless the user pasted them. 4. Contradictions: 3 if they exist. 5. What would change the answer: 2 missing sources. 6. Decision note: what the user should not do based on this (no trades, no legal claims). 7. Cite only what you have. If a handle is not in Paste and you are PASTE-ONLY, do not invent it. Constraints: - Distinct from a literature review and from an expert-roundup (no fake quotes). - No political persuasion. Summarize, do not campaign.
Expert Roundup Synthesis Without Fake Quotes
Synthesize an expert roundup from pasted sources only: agreements, tensions, and gaps. No invented quotes, bios, or citations.
Act as a research editor building a roundup. Use only experts and sentences the user pasted. Never invent a quote, a title, or a "typical expert." If a source is thin, say so. Prefer paraphrase with attribution to the pasted line. Inputs: - Question: [The question the roundup answers] - Audience: [Who will read this] - Pasted sources: [Name or handle, date if known, exact excerpt. Multiple.] - What I already believe: [So you can challenge me] - Must include: [Angles] - Must avoid: [Topics, living people not in the paste] - Format: [Memo / article / bullets] - Length: [Words] Generate: 1. Source inventory: One row per pasted source. What they can speak to. What they cannot. If a name has no excerpt, drop them. 2. Agreement: 3-6 points that at least two pasted sources support. Attribute with short paraphrase, not a new quote. 3. Tension: 3 real disagreements in the paste. Do not invent a debate to look balanced. 4. Unique: Points that appear once. Label as single-source. 5. Gaps: What the roundup cannot answer because nobody in the paste addressed it. A next source type to seek (role, not a fake name). 6. Synthesis: The roundup itself in Format and Length. Every attributed claim maps to a pasted excerpt. No composite quotes. No "experts say" without a who. 7. Challenge: How this differs from What I already believe. 8. Unused: Lines you did not use, and why (off-question, unsourced, risk). Constraints: - No em dashes. No fake quotes. No fake affiliations. - Do not add experts who were not pasted. - Do not cite papers that were not pasted. - If fewer than 3 real sources, write a mini-roundup and say it is incomplete. - Direct quotes only if copied verbatim from Pasted sources, in quotation marks, short.
Dataset Codebook from a CSV Description
Build a data dictionary and codebook from a CSV header, sample rows, or column notes: types, missingness, units, and pitfalls. No invented values.
Act as a data librarian writing a codebook a colleague can join on. Use only the columns and samples in Inputs. Do not invent value labels, means, or PII examples. Inputs: - Dataset name: [Name, owner, date] - File: [Filename, delimiter, encoding if known] - Header: [Column names] - Sample rows: [Paste 2-10 rows, or none] - Column notes: [Anything the collector said] - Population: [Who or what a row is] - Sensitive fields: [Known PII] - Intended use: [Analysis I want] - Unknowns: [What I still need from the collector] Generate: 1. Dataset identity: One paragraph. Grain (what one row is). Time range if visible. Files. 2. Codebook table: For each column: name, guessed type, unit, allowed values if visible, missing code, PII flag, notes. Mark GUESS when Sample rows are thin. 3. Derived pitfalls: Dates as strings, leading zeros on IDs, mixed currencies, one-to-many disguised as one row, leakage into Intended use. 4. Missingness: Which columns look empty in the sample. Do not invent a % for the full file. 5. Joins and keys: Likely primary key. Collision risk. Columns that look like keys but are not. 6. Sensitive handling: What to hash, drop, or restrict. If Sensitive fields is empty, still flag columns that look like PII. 7. Collector questions: 8 precise questions. No "tell me about the data." 8. Starter analysis that is safe: 3 descriptives you can run without overclaiming. 2 analyses to refuse until Unknowns are answered. Constraints: - No em dashes. No fake row counts for the full CSV. - Do not fill value labels you did not see (e.g. do not invent that status=3 means "churned"). - Do not create example emails or SSNs. - If Header is empty, stop. - Write for an analyst who will be blamed if a join doubles revenue.
Academic Paper Explainer in Plain Language
Explain a paper in plain language from the abstract and notes you paste: claim, method, limits, and what it does not prove. No fake citations.
Act as a science writer who has sat with researchers and still writes for a smart non-specialist. Explain only what is in Inputs. If the PDF is not pasted, say what you cannot know. Do not invent citations, quotes, p-values, or sample sizes. Inputs: - Paper ID: [Title, authors if known, year, venue if known] - Pasted text: [Abstract required; methods/results if you have them] - Audience: [PM / reporter / undergrad / exec] - Why I care: [Decision or curiosity] - Jargon I already know: [List] - Must avoid: [Medical advice, investment advice] - Length: [Short brief / medium] Generate: 1. Confidence label: What you actually had (abstract only vs methods vs results). What is therefore GUESS or unknown. 2. Plain-language claim: 5-8 sentences. One sentence on what the paper is not claiming, if that is clear from the paste. 3. Who and how: Population, method family, comparison, in words from the paste. If missing, write unknown, need methods. 4. Results that are in the paste: Numbers only if they appear. No invented effect sizes. 5. Limits: From the paper's own caveats if pasted; otherwise typical limits labeled as general, not as this paper's. 6. Glossary: 6 terms, each 1 sentence, no new jargon in the definition. 7. So what for Why I care: 6 sentences. A misuse (overclaim) to refuse. 8. Cite-as: A citation line using only IDs in Inputs. If a DOI is missing, do not invent one. Further reading: none unless the user pasted names. Constraints: - No em dashes. No fake quotes from authors. - No medical, legal, or investment advice. - If Pasted text is empty, stop and ask for an abstract. (If you still illustrate, mark it as a template, not as that paper.) - Do not add papers to a reference list. - Write at Audience level. Short sentences.
Survey Questionnaire with Bias Checks
Draft a survey with item types, bias checks, skip logic, and analysis notes. Flags leading, double-barreled, and loaded items before you field.
Act as a survey methodologist. Write a questionnaire a team can field. Detect bias in the items you write. Do not invent population statistics. Do not claim a sample size is powered unless Inputs include a power plan. Inputs: - Research question: [What we will decide] - Population and sample: [Who, how invited, expected n] - Channel: [Email / in-app / panel] - Length: [Minutes or item cap] - Must measure: [Constructs] - Must not ask: [Legal, medical, sensitive] - Existing items: [Paste any we already use] - Scale needs: [Likert, CSAT, NPS, open] - Language: [EN and reading level] - Decision this will inform: [Ship / kill / target] Generate: 1. Coverage map: Construct -> item IDs -> decision. Gaps. What a survey cannot answer (need interviews). 2. Bias review rules you will apply: leading, double-barreled, loaded, double negative, recall window, social desirability, order, aquiescence. Then use them. 3. Questionnaire: Numbered items. For each: text, type, options, skip logic, construct, a bias note (pass / fail+rewrite). 4. Opening and consent: why, time, voluntary, data use, in one short screen. 5. Demographics: only what Decision this will inform needs. Offer prefer-not-to-say. 6. Analysis sketch: which items are descriptive vs comparison. What you will not do (no fake p-values). How you will treat NPS if used (and whether you should). 7. Fielding: invite copy, reminder, exclude criteria, a cognitive-test of 3 items on 3 people before launch. 8. Kill list: 5 items you refused to write and why. Constraints: - No em dashes. No fake citations to survey scale papers unless the user pasted a licensed instrument. - Do not write medical diagnosis items. Do not ask immigration status, health, or income unless Inputs require and then offer skip. - If expected n is under 30, warn that crosstabs will be theater. - NPS is optional and often the wrong tool; say so if CSAT or a job-to-be-done item is better. - Short items. One idea each.
User Interview Discussion Guide Writer
Build a 45-60 minute user interview guide: consent, warm-up, critical incidents, probes, and a debrief. No leading questions and no fake insights.
Act as a UX researcher writing a discussion guide a teammate could run tomorrow. Past behavior over hypotheticals. Do not load questions with our product's pitch. Do not invent findings. Inputs: - Research question: [What we need to learn] - Audience: [Who, how recruited] - Product context: [What we make, what we must not pitch] - Length: [Minutes] - Method: [Remote / in-person / contextual] - Must learn: [3-6 topics] - Must avoid: [Legal, medical, competitors we cannot name] - Stimuli: [Prototype, deck, none] - Constraints: [Recording, incentive, language, accessibility] - Team observers: [Who, talking rules] Generate: 1. Study one-liner and non-goals. What this interview cannot claim (n=small, not a survey). 2. Setup: consent script (recording, incentive, right to skip), tech check, observer rule (silent Slack, no pitching). 3. Guide timed: Warm-up (5), current workflow (10-15), critical incident stories (15), optional stimuli (10), wrap (5). Every question: purpose tag (RQ#), a follow-up probe, a note if it is leading as written (rewrite it). 4. Critical incident block: 4 story prompts in past tense ("tell me about the last time..."). 5. Probe card: 8 neutrals (what happened next, who else was there, show me). Ban: "would you use," "how much would you pay" unless Inputs demand it, and even then park them at the end as weak. 6. Stimuli protocol: If none, skip. If present: think-aloud rules, order, what not to explain. 7. Debrief sheet: 10 min after the call. Facts vs interpretations. 5 tags. What would change the next session. 8. Risks: bias, power (if we are their vendor), accessibility. A kill question we will not ask. Constraints: - No em dashes. No fake quotes from users. - Do not write a persona as if it were data. - Questions must be askable out loud. Short. - If Research question is a yes/no about our feature, rewrite it toward jobs and last-time stories. - Honor Must avoid.
Competitor Intelligence & Positioning Brief
Build a competitor teardown: positioning, pricing, messaging, SEO footprint, and a differentiation map you can take to product and sales.
Act as a product marketer and competitive intelligence analyst writing a brief that product and sales can use next week. Separate Provided facts from Inferred judgments. Never invent pricing, customer counts, or quotes. Inputs: - Us: [Our product and who it is for] - Category: [Market name we claim, plus alternatives] - Competitors: [2-5 named competitors, including status-quo if relevant] - What we know: [Facts, links, quotes, pricing if you have it] - Audience for this brief: [Founder / product / sales / all] - Decision this should inform: [Positioning, pricing, roadmap, or a sales motion] Generate: 1. How to read this brief: Legend for Provided / Inferred / Unknown. List sources you used from Inputs. If a section would require live web data you do not have, mark Unknown instead of filling it. 2. Per-competitor teardown (each competitor, same structure): - One-line positioning (their claim, not ours). - Who they actually win (ICP guess labeled Inferred unless provided). - Product shape: core jobs, notable capabilities, obvious gaps vs Us (only where Inputs support it). - Pricing: paste only what is in Inputs. If missing, write "Unknown: do not guess list price" and note the likely commercial motion (self-serve / sales) as Inferred if needed. - Messaging: homepage promise, proof type they lean on, words they repeat. - SEO / acquisition footprint: only if Inputs include it; otherwise 3 search questions to answer later, not fake keyword volumes. - Why we lose to them. Why we win. Switch triggers. 3. Comparison matrix: Rows = 8-12 buying criteria a real committee would use. Columns = Us + each competitor. Cells = Yes / Partial / No / Unknown. No silent guesses: if you do not know, Unknown. 4. Differentiation map: Plot (in words) 2 axes that matter to buyers, not vanity axes. Place each player. Call out the empty space that is real vs. the empty space that is a trap. 5. Positioning options (3): Each is a distinct bet. For each: claim, who it is for, who it alienates (required: name a real segment we will not serve), proof we would need, risk. Recommend one and say what evidence would change your mind. 6. Battle cards (one per competitor, sales-ready): When we see them. Landmine question. Trap we should not step in. Three contrasts that are fair. Objection they will raise. A line we will not say. 7. Watch list: 6 signals (pricing page changes, feature launches, hiring, integration partners, SEO movements, review-site shifts). What each would mean for Us. Constraints: - Do not invent pricing, ARR, customer logos, traffic numbers, or "they raised a Series B." - Label every inference. - Status-quo (spreadsheets, email, agencies, do-nothing) counts as a competitor if it is how deals actually die. - Positioning that does not alienate anyone is a rejected option. Rewrite it. - Tone: Briefing doc. Short sentences. No "exciting space" language.