Your job, on Claude.

You're my co-editor on [BOOK TITLE], a [genre] novel. Everything you need is in this Project's knowledge: my style guide, the story bible, and the chapters written so far. Before answering any question in this Project, check those files — the bible outranks your assumptions. Start by reading everything and giving me back a one-page summary of the book as you understand it: premise, cast, timeline, open plot threads, and the three stylistic rules you'll be most careful to protect. Then flag anything in the bible that's contradictory or underspecified, so I can fix it now instead of discovering it in chapter twenty.

Act as my developmental editor. I've uploaded my draft — do not rewrite anything yet. Give me a structural edit letter: (1) what this book is actually about, in one sentence based only on the pages — I want to see if it matches my intent; (2) chapter by chapter, what each is doing, where it drags, and what's missing; (3) the three biggest structural problems, ranked by damage; (4) what I should cut entirely; (5) the strongest 10% I should protect at all costs. Be specific — quote lines and cite chapters for every claim. Grade the structure honestly, like an editor whose name goes on the book.

Read every chapter in this folder in order and build a series bible as a new file: every character (physical details, relationships, first appearance, speech habits), the timeline with dates and elapsed time, locations, and world rules — each entry with the chapter reference where it's established. Then run a continuity audit of the newest draft against that bible: contradictions, dropped subplots, characters whose eye color or age drifted, timeline math that doesn't add up. Deliver the audit as a second file, ordered by severity, with chapter and line references for every hit. Don't fix anything — I decide what's canon.

Here's my outline for [chapter / issue / essay] — each section lists the point it must land. Draft it in MY voice using the style guide and samples in project knowledge. Follow documented patterns, not vibes: match my sentence-length rhythm, my openings, the words I never use, and everything on the style guide's 'never' list. When you finish, grade your own draft against the style guide section by section — flag every line that sounds more like an AI than like me, and rewrite the flagged lines before showing me anything. Deliver the draft plus a short list of places you deviated from the outline and why.

Every Monday at 7am: you're the research desk for my [book on X / newsletter about Y]. Search the web for what's new in the last seven days — studies, credible reporting, notable arguments, useful data points, and anything that contradicts positions I've already published. Write a digest document: each item gets a two-sentence summary, why it matters for my project, and the source link — real, clickable URLs only; skip anything you can't link. End with three angles worth writing about this week and one 'someone is wrong on the internet' opportunity. Keep it under two pages, and if the week was genuinely quiet, say so instead of padding.

Take chapters 1–3 (attached) and produce two files. First, a submission-standard manuscript .docx: Times New Roman 12pt, double-spaced, 1-inch margins, header with [SURNAME / TITLE / page number], each chapter starting on a new page one-third down. Second, a print-interior PDF at 6x9 inches with mirrored margins, running heads, and front-matter placeholders (title page, copyright, dedication). Formatting only — don't touch the prose, but flag anything you had to normalize, like scene-break marks or em-dash spacing. Before handing them over, open your own output and check page size, margins, headers, and orphaned chapter titles; fix what fails, then give me both download links.

You're a web designer building my personal portfolio as a single-page HTML artifact. I've uploaded my resume, descriptions of my projects, and screenshots of two sites whose design I admire. Build the complete page: a hero with my name and one-line positioning, selected projects with outcomes, a short about section, and a contact block. Design direction: sophisticated and editorial — generous whitespace, a restrained two-color palette, real typographic hierarchy, and no AI-looking gradients, emoji headings, or cookie-cutter card grids. Write the copy from my actual materials; don't invent clients, metrics, or testimonials. When you think you're done, grade your own work against my inspiration screenshots on layout, typography, and personality, and iterate until it would pass for a hand-built site.

I write a personal-finance blog and I'm publishing a post on emergency funds. Build a self-contained interactive calculator as an HTML artifact that I'll embed in the post: readers enter monthly essential expenses, income stability (dropdown: steady salary / variable / freelance), and number of dependents, and it recommends a 3–9 month savings target with a visual progress bar and one plain sentence explaining the recommendation. Design: clean and calm against a white blog background, system fonts only, no external images or fonts — everything must work with zero network access. Mobile-first; most of my readers are on phones. Include a friendly line that this is guidance, not financial advice. Before showing me, test it yourself as three different readers — a freelancer, a two-income family, a new grad — and fix anything confusing.

Using the style guide and past posts in this project's knowledge, draft a 1,400-word post: "How long does sourdough starter really take?" Target reader: a first-time baker who tried once and gave up. Structure it for search: an H1 that matches the question, a direct two-to-three sentence answer at the top, H2s for the sub-questions people actually ask, an FAQ section, a meta description under 155 characters, and three title options. Match my voice from the samples — short paragraphs, first person, one dry joke maximum, no listicle filler. Where a claim needs a number or source, mark it [VERIFY] instead of guessing. End with two internal-link suggestions pointing to my existing posts in project knowledge. Then critique your draft against my style guide and revise once before showing me.

You're a conversion copywriter auditing my blog's core pages. I've pasted my homepage, about page, and three most-read posts. Audit them as a first-time visitor: (1) Can you tell within five seconds who this site is for and what they get? (2) Flag jargon, filler, and sentences doing no work — quote them exactly. (3) Check every call-to-action: where it appears, whether it's specific, what's missing. (4) Note tone drift — where do I stop sounding like the same person? (5) Find outdated references: years, prices, dead projects, old job titles. Deliver a table with page, issue, severity, quoted text, and a suggested rewrite. Then give me the three highest-impact fixes if I only have one evening. Be blunt — polite feedback wastes both our time.

I publish a post tomorrow: "A beginner's guide to backyard composting." You don't generate photos, so do every job around the image instead. (1) Write three detailed image-generation prompts I can paste into Midjourney or Canva — subject, composition, lighting, mood, 16:9, no text in the image. (2) Give me three search phrases each for Unsplash and Pexels likely to surface strong free photos, plus what to avoid (sterile stock smiles, obvious AI gloss). (3) Draw one simple SVG diagram of compost layering — browns, greens, moisture — as an artifact, in a green-and-cream palette that suits a warm gardening blog. (4) Write alt text and a one-line caption for all four visuals. Keep the set visually consistent, and tell me which single visual matters most if I only have time for one.

Every Monday at 7am: I write [your blog] about [your niche] for [your audience]. Research what happened in this niche over the last 7 days — notable articles, discussions, product or policy changes — and check what these sites published: [your list]. Then give me a one-page briefing: (1) three post ideas ranked by fit for my audience and how thin the existing coverage is, each with a working title, a five-bullet outline, and the angle nobody is taking; (2) one older post of mine worth updating this week, and why; (3) two topics everyone else covered that I should skip, with reasons. Don't draft full posts — the briefing is the deliverable. If a claim from a source looks shaky, say so instead of repeating it.

I run paid ads for a small e-commerce brand. Attached is our Q2 export from Meta Ads Manager with spend, impressions, clicks, and conversions by campaign, ad set, and ad. Context you need: target ROAS is 3x, new-customer lifetime value is roughly double the first order, and I can shift up to 30% of budget between campaigns next quarter. Analyze this like a senior performance marketer: which campaigns and audiences beat target, which quietly burned money, and where creative fatigue shows up in the week-by-week numbers. Then recommend a specific Q3 budget reallocation with expected impact. Deliver a short written summary I can paste to my boss, plus an Excel file with the supporting tables and charts. Grade your own analysis before you finish — if a recommendation isn't backed by a number in the export, cut it.

Here's this week's export from our ad account (CSV attached). Build me an interactive dashboard artifact: a top row of four KPI tiles — Total Spend, CPA, ROAS, Conversions — each showing change vs the prior period. Below that, a spend-versus-return chart by campaign, and a table of the five best and five worst ads by CPA. Design it clean and calm, like a well-set report page, not a cockpit — no neon gradients. Bake this week's numbers directly into the artifact. Every week I'll paste the new export into this same conversation and say "refresh" — when I do, keep the layout exactly the same and update only the numbers and the week-over-week deltas.

Using this Project's knowledge — our brand voice, our ICP, and the winning ads with their performance numbers — build a creative test matrix for our spring launch campaign. I want 10 hooks across 3 distinct angles (pain-led, proof-led, curiosity-led), each written at platform length for Meta primary text plus a six-word headline version. For every variant, add one line of hypothesis: what it tests and why our past winners suggest it could work. Lay it out as a table I can drop straight into our test plan. Then critique your own matrix: kill any two variants that are really the same idea in different clothes and replace them with something genuinely different.

I'm rebuilding our customer personas from primary evidence, not assumptions. Attached: six months of support tickets, our post-purchase survey export, and notes from twenty sales calls. Cluster actual behavior and language into three or four personas. For each one: who they are, what job they're hiring us for, their exact words (quote the files directly), where they get stuck, and what would make them leave. Flag anything in the data that contradicts our current assumptions — that's the most valuable part. Present it as an interactive artifact I can click through persona by persona, with a journey map and the real quotes pinned to each stage. If the evidence only supports two personas, give me two — don't invent a third to make it feel complete.

Every Monday at 7am, research what changed last week for these competitors: [Competitor A], [Competitor B], [Competitor C]. Check their websites for pricing and product changes, search news and recent coverage for launches, funding, or leadership moves, and scan public conversations — Reddit threads, review sites — for shifts in how customers talk about them. Also flag notable movement on our category topics: [topics]. Write it as a five-minute briefing: what changed, why it matters to us, and one suggested response for each item, with source links so I can verify. If nothing meaningful happened, say so in two lines — don't pad. Leave the briefing as a document I'll find with my coffee.

I want to create a skill for my brand guidelines. Interview me about everything you'd need to apply them without being reminded: our color palette (I'll give you hex codes), typography, logo usage, voice and tone rules — including words we never use — and how all of it differs between ads, emails, landing pages, and decks. Then build the skill so that any time I ask for a deliverable, you automatically write and design it on-brand, and you flag anything that breaks a rule instead of silently shipping it. Before you finish, test your own work: generate a sample ad and a sample slide, grade them against the guidelines, and fix whatever misses.

You're my sales research analyst. I'm an account executive selling [your product] and I have a first call with [company] next week. Build me an account brief I can absorb in ten minutes: what the company does and who they sell to, headcount and any recent funding or leadership changes, news from the last six months, tools or vendors they publicly use, and three plausible business pains our product could map to. Cite your sources so I can verify anything I quote out loud. Then present it as a clean one-page artifact — company snapshot at the top, "why now" signals in the middle, suggested discovery questions at the bottom. Before you hand it over, grade the brief against "would a VP of Sales find this credible?" and rewrite anything that reads as generic filler.

You're my chief of staff. Look at my Google Calendar for tomorrow and find every external meeting. For each one, pull the matching contact and deal from HubSpot — stage, amount, last activity, and any notes — then search the web for news about their company from the past month. Give me one brief per meeting: who I'm meeting and their role, where the deal stands, what happened last time we talked, one relevant piece of recent news, and the single most important thing to accomplish on this call. Flag any meeting where the CRM record is thin or the deal has been silent for more than 21 days. Keep each brief under 150 words — I'll be reading these between calls, not at a desk.

You're my deal-savvy sales assistant. Here are my raw notes from today's call with [name] at [company]: [paste notes]. Draft the follow-up email. Structure: one line of genuine specificity about the conversation (not "great chatting"), a three-bullet recap of what we agreed, the one resource I promised, and a clear next step with a proposed date. Tone: warm, direct, zero filler — a good rep who respects their reader's inbox. Under 150 words. Then write a second, shorter version my champion can forward to their boss — that one has to survive being read in six seconds. If my Gmail is connected, save the first version as a draft so it's ready when I am.

Using the past proposals and pricing in this Project as your model for structure and voice, draft a proposal for [company] — [a paragraph of deal context: what they need, deal size, timeline, who's deciding]. Follow our standard arc: executive summary written in their language (their problem, not our product), proposed solution mapped to the three pains from discovery, implementation timeline, pricing pulled from the current sheet only, and the most relevant case study from the Project knowledge. Match the tone of our past winners — confident, concrete, no adjectives doing work the numbers should do. Deliver it as a formatted Word document. Before you hand it over, grade it against our best past proposal and rewrite the weakest section.

Roleplay with me. You are the CFO of a 150-person logistics company evaluating [your product] at [price]. You're skeptical, budget-conscious, and you've been burned before by a tool that promised ROI and delivered a login screen. Your real objections: the price versus your current spreadsheet workflow, the integration effort, and "why now." Stay in character — push back hard, answer weak responses with sharper follow-up questions, and don't soften just because I flounder. We'll go ten exchanges. Then break character and grade me: which of my responses would have actually moved a real CFO, which were filler, and what is the one reframe I should practice this week? Be blunt — a kind score sheet teaches me nothing.

Every Monday at 7am, review my HubSpot pipeline and write my meeting memo. Pull all open deals: group them by stage with totals, then flag the three things I always miss — deals with no activity in 14+ days, deals past their expected close date, and deals missing a next step. Compare total pipeline to last week and say plainly whether coverage is up or down, and why. Close with three recommended actions for the week, ranked by revenue at stake. Format it as a one-page memo I can read before the 9am meeting: numbers first, narrative second, no pep talk. If anything looks like a data problem rather than a deal problem — a $0 amount, a close date in the past — call that out in its own section.