Manus Prompt Templates (2026): Fill-in-the-Blank

GOAL: Research and write a report on [topic / question] for [audience, e.g. a product team deciding whether to enter this market].
SCOPE: Cover [3-5 specific sub-questions]. TIMEFRAME: only sources from [date range, e.g. Jan 2024 onward]. GEOGRAPHY: [region]. Exclude [what to leave out].
SOURCES: Prioritize [primary sources / official docs / named sites]. Avoid [low-quality sources]. Use the Browser Agent to open and read the actual pages, not just search snippets.
DELIVERABLE: A [length, e.g. 1,500-word] report with these sections: Summary, [section 2], [section 3], [section 4], and Recommendations. Add a table of [key comparison, e.g. options vs. criteria].
AUDIT: Cite every claim with a source URL. Leave a field blank rather than guess. Tag any uncertain claim with a confidence level (High/Med/Low). End with a sources appendix listing every URL. If any of the scope or sources is unclear, ask me before you start.

GOAL: Research each item in this list and build a comparison table: [paste list of 50-200 items — companies, products, papers, or tools].
SCOPE: For every item, find [attribute 1], [attribute 2], [attribute 3], and [attribute 4]. TIMEFRAME: current as of [date].
METHOD: Use Wide Research to work the list in parallel — one sub-agent per item — and the Browser Agent to read each item's real page.
DELIVERABLE: A spreadsheet with one row per item and columns: Name, [attribute 1], [attribute 2], [attribute 3], [attribute 4], Source URL, Confidence.
AUDIT: Cite the source URL for every cell that states a fact. Leave a cell blank rather than guess, and mark it "not found". Tag each row's overall confidence High/Med/Low. Add a notes column for anything ambiguous.

GOAL: Produce a market report on [market / category] in [geography] for [audience].
SCOPE: Cover market size and growth, the top [number] players, pricing models, key trends, and [regulatory / risk factor]. TIMEFRAME: [year range].
SOURCES: Use industry reports, company sites, and reputable news — cite each. Use Wide Research to profile the top players in parallel. Avoid vendor marketing claims stated as fact.
DELIVERABLE: A [length]-word report with sections: Overview, Market Size, Competitive Landscape (with a table of players vs. [criteria]), Pricing, Trends, Risks, and Outlook.
AUDIT: Every figure and claim gets a source URL. Leave a field blank rather than estimate, or clearly label any estimate as an estimate. Tag forward-looking claims Med or Low confidence. Sources appendix at the end.

GOAL: Do a deep dive on [single subject — a company, technology, person, or product] for [audience].
SCOPE: Answer specifically: [question 1], [question 2], [question 3]. Go past the surface — read primary sources, filings, docs, and first-hand accounts, not just top search results.
SOURCES: Use the Browser Agent to open [named primary sources] directly. TIMEFRAME: emphasize the last [period], but include key history.
DELIVERABLE: A structured brief with: TL;DR (5 bullets), Background, [the specific questions as headed sections], What's uncertain, and Sources.
AUDIT: Cite every claim with a source URL. Where sources disagree, present both and say which is better supported. Leave gaps explicit rather than filling them. Tag uncertain points High/Med/Low.

GOAL: Build a list of [number, e.g. 50] prospect companies matching my ICP.
ICP / SCOPE: [industry], [company size / employee range], [geography], [other filters, e.g. uses X technology, raised funding since Y]. Exclude [what to skip, e.g. current customers, competitors].
METHOD: Use Wide Research to find and qualify candidates in parallel, and the Browser Agent to confirm each against the ICP from its real site.
DELIVERABLE: A spreadsheet with columns: Company, Website, [industry], [size], [geography], Why it fits, Source URL.
AUDIT: Only include a company you can verify against a source — cite the URL. Leave a field blank rather than guess. Flag any borderline fit with a Med/Low confidence tag. STOP when you have [number] verified rows.

GOAL: Enrich this list of [companies / people] with the fields below: [paste list].
SCOPE: For each row, find [field 1, e.g. company size], [field 2, e.g. HQ location], [field 3, e.g. relevant decision-maker + role], [field 4, e.g. recent news or trigger event]. TIMEFRAME: current.
METHOD: Use Wide Research to enrich rows in parallel and the Browser Agent to read each company's real site and public profiles. Do not scrape anything behind a login.
DELIVERABLE: The same list returned as a spreadsheet with the new columns added, plus a Source URL and Confidence column per row.
AUDIT: Cite a source URL for every enriched cell. Leave a cell blank and mark "not found" rather than guessing. Tag each row High/Med/Low. Note any row where identity was ambiguous.

GOAL: Analyze this dataset and answer [the business question]: [attach / paste data or link].
SCOPE: Focus on [metrics / dimensions that matter]. Segment by [dimension, e.g. region, month, product]. TIMEFRAME: [period covered].
METHOD: Write and run code to clean and analyze the data. Note any data-quality issues (missing values, duplicates, outliers) before drawing conclusions.
DELIVERABLE: A short report with: the direct answer, the top [number] findings each backed by a number, a table of [key breakdown], and [chart(s), e.g. trend over time]. Return the cleaned data file too.
AUDIT: Show the calculation or filter behind each headline number so I can reproduce it. Flag any finding that rests on thin or messy data with a Low confidence tag. State every assumption you made about the data.

GOAL: Build an interactive dashboard from this data for [audience] to track [what they care about]: [attach / paste / link data].
SCOPE: Show these metrics: [metric 1], [metric 2], [metric 3]. Allow filtering by [dimension]. TIMEFRAME: [period]. Refresh logic: [static snapshot / describe how to update].
DELIVERABLE: A working dashboard (web page) with KPI tiles for the headline metrics, [chart types] for the trends, and a filterable table. Clean, mobile-friendly layout. Include a one-paragraph "how to read this" note.
AUDIT: Show the definition and source column behind every metric so the numbers are auditable. Flag any metric derived from incomplete data. Note any assumption in a visible footnote on the dashboard.

GOAL: Create a [number]-slide deck on [topic] for [audience], to be used for [context, e.g. a sales pitch / internal review].
SCOPE: Cover [the key points, in order]. TONE: [tone]. Research any facts you need from [sources] first.
DELIVERABLE: Use Manus Slides. Structure: title slide, [problem/context], [key findings — one point per slide], [implications], and a clear next-steps slide. Each content slide gets a headline and 3-4 tight bullets. Add a final Sources slide.
AUDIT: Put a source citation on any slide that states a fact or figure — leave nothing unsourced. Flag any slide where the supporting evidence is thin. Do not invent statistics; if a number isn't in the sources, leave the slide qualitative and say so.

GOAL: Write a one-page [report / brief / memo] on [topic] for [audience] who need to [decision or action].
SCOPE: Answer [the core question] and cover [2-3 must-hit points]. Keep it to one page — cut anything that isn't decision-relevant. TONE: [tone].
SOURCES: Base it on [attached material / researched sources]. Cite as you go.
DELIVERABLE: A clean one-pager: a one-line bottom-line-up-front, 3-4 short sections with headers, and a single recommendation. Include a small table of [key comparison] if it helps.
AUDIT: Cite every fact with a source URL. Leave out anything you can't support rather than padding. Tag any claim you're unsure of with a confidence level and list open questions at the bottom.

GOAL: Build a responsive landing page for [product / offer] aimed at [audience].
SCOPE: The key message is [value proposition]. Primary call to action: [CTA]. Include sections: hero (headline + subhead + CTA), [benefit blocks], social proof placeholder, FAQ, and a closing CTA. TONE: [tone]. Brand colors: [colors].
CONTENT: Write the copy too — headline, subhead, benefit bullets, and FAQ answers — grounded in [what the product actually does / attached notes]. Do not overclaim.
DELIVERABLE: A working, mobile-friendly page I can preview and download. Clean semantic HTML/CSS.
AUDIT: Do not invent testimonials, stats, or customer names — use clearly labeled placeholders for anything I need to supply. Flag any claim in the copy that needs legal or factual review before it goes live.

GOAL: Build a simple web tool that [does what — e.g. calculates X from user inputs / lets me look up Y].
SCOPE: Inputs: [fields the user enters]. Logic: [what it computes or does, step by step]. Output: [what it shows]. Keep it single-purpose — no accounts, no backend unless required.
DELIVERABLE: A working web app I can open in the browser and use immediately, with a clean, mobile-friendly interface and clear labels. Include short in-app instructions.
AUDIT: Test the tool with [2-3 example inputs] and show the results so I can confirm the logic is right. Handle empty or invalid input gracefully. Document any assumption you made about the logic in a visible note, and list anything you'd want me to confirm.

GOAL: Set up a Scheduled Task that monitors [what — e.g. these 5 sources / this topic / these pages] and reports anything new.
CADENCE: Run [every weekday at 8am / weekly on Monday]. SOURCES: Check [list the exact sites, feeds, or searches]. Use the Browser Agent to open each and read the real page.
WHAT COUNTS: Report only items that are [genuinely new since last run / match these keywords: ...]. Ignore [noise to skip].
DELIVERABLE: Append new items to [a running spreadsheet / send a short digest] with columns: Date, Item, Why it matters, Source URL. If nothing new, say "nothing new" — don't manufacture updates.
AUDIT: Cite the source URL for every item. Never repeat an item already reported. Leave "Why it matters" blank rather than guess if it's unclear.

GOAL: Set up a Scheduled Task that tracks my competitors and flags meaningful changes.
CADENCE: Run [weekly on Monday]. COMPETITORS: [list of 3-10 competitors + their key pages — pricing, product, blog, careers]. Use the Browser Agent to check each page.
WHAT COUNTS: Flag changes to [pricing, new features/launches, messaging, hiring signals, funding/news]. Ignore trivial edits like typos or date changes.
DELIVERABLE: Append to a tracker spreadsheet with columns: Date, Competitor, What changed, Category, Why it matters, Source URL. Add a one-line summary of the week at the top.
AUDIT: Cite the exact page URL for every change. Only report a change you can verify against the live page. Tag the significance of each change High/Med/Low. Never invent a change to fill the report.

GOAL: Set up a Scheduled Task that produces a weekly digest on [topic / area I follow] for [me / my team].
CADENCE: Run every [Friday at 3pm]. SOURCES: Pull from [named newsletters, sites, feeds, or searches]. Use the Browser Agent to read the actual articles, not just headlines.
SCOPE: Cover the [number] most important developments of the week in [topic]. Skip [what to leave out]. Keep it skimmable.
DELIVERABLE: A short digest: a 2-sentence overview, then [number] items each with a headline, a 2-line "why it matters", and a source link. Group by [theme if useful].
AUDIT: Cite a source URL for every item. Include only things you can verify from a real source. If it was a quiet week, say so and keep it short rather than padding with filler.