Claude Prompts

You are my front-end developer. Build an interactive Artifact: a single self-contained HTML file with inline CSS and JavaScript, no external dependencies except one CDN chart library. It must serve as a [DASHBOARD PURPOSE, e.g. freelance income tracker] for [USER TYPE]. Requirements: [NUMBER] KPI cards, [NUMBER] charts (types: [CHART TYPES]), a filter control for [FILTER DIMENSION], and a data table with columns [COLUMNS]. Design: [STYLE, e.g. clean light theme], mobile responsive, and every text label editable by clicking it. Prefill it with realistic sample rows for [SCENARIO]. The Artifact must work the moment it renders, with no setup steps. After the code, explain in 5 bullets how to customize the data source and which lines to change to add a new chart.

Act as my knowledge manager. I am creating a Claude Project for [PROJECT NAME] about [DOMAIN]. Draft: 1) the project description and custom instructions (voice: [TONE]; always cite which knowledge-base document an answer came from; if the answer is not in the knowledge base, say so and offer to search the web), 2) a list of [NUMBER] documents to upload (for example [DOC TYPES]) with a one-line note on what each contributes, 3) 5 starter prompts the team ([TEAM SIZE] people in [ROLES]) will use most, 4) rules for what belongs in the Project versus what stays in a normal chat. Keep the instructions under [WORD LIMIT] words and make them enforceable, not aspirational.

You are directing Claude Code on my codebase. The repository is [REPO DESCRIPTION, e.g. a Next.js app with a Postgres database] and the task is [TASK, e.g. add rate limiting to the public API routes]. Write the briefing Claude Code will execute: 1) the goal in one paragraph plus the definition of done ([NUMBER] acceptance criteria), 2) the files and folders it must read first ([PATHS]), 3) constraints: do not modify [PROTECTED AREAS], follow the existing [STYLE GUIDE OR PATTERNS], keep changes minimal, 4) the verification steps it must run (commands: [COMMANDS, e.g. npm test, npm run lint]) and what passing looks like, 5) what to do if a test fails ([FAILURE PROTOCOL]). Write it as instructions Claude Code can execute autonomously, with checkpoints where it reports back to me.

You are my research analyst. I will paste or upload a long document ([DOC TYPE, e.g. a 150-page industry report]) about [TOPIC]. Once I share it, do the following in order: 1) confirm the document length, structure, and date, and flag any sections that seem missing, 2) produce a [WORD COUNT]-word executive summary focused on [ANGLE OF INTEREST], 3) extract [NUMBER] key findings as a numbered list, each with its page or section reference, 4) build a comparison table of [ENTITIES, e.g. the 5 vendors reviewed] across [CRITERIA], 5) list [NUMBER] questions the document raises but does not answer. Do not start until I say the words [START PHRASE, e.g. document uploaded]. Quote directly, with section references, for any claim I might need to defend to [STAKEHOLDER].

You are my writing coach. I want a custom Claude Style that writes like [STYLE REFERENCE, e.g. our company blog]. The style must use a [TONE WORDS] tone, average [NUMBER] words per sentence, open pieces with [OPENING PATTERN], structure text with [STRUCTURE, e.g. short paragraphs and subheadings every 200 words], and avoid [AVOID LIST]. To calibrate it, I will paste [NUMBER] samples of writing I like. After I paste them, analyze the samples and write the Style definition as: 1) a 3-sentence style summary, 2) 8 concrete rules written as do and don't pairs, 3) a before-and-after rewrite of my draft paragraph about [TOPIC] showing the style applied. Then ask me which rule to adjust first.

Act as my fact-checker. Here is a claim I plan to publish: [CLAIM]. It will appear in [PUBLICATION CONTEXT, e.g. a client proposal]. Search the web and return: 1) a verdict (Supported, Partly supported, Unsupported, or Outdated), 2) the [NUMBER] strongest sources with publication dates and one line on why each is credible on this topic, 3) the exact numbers or quotes that confirm or contradict the claim, 4) a rewritten version of the claim that is fully defensible, under [WORD COUNT] words, suitable for [AUDIENCE]. If the evidence is thin or comes from a single source, say so clearly and suggest [NUMBER] ways I could verify further ([VERIFICATION IDEAS]).

You are my data analyst. I will upload [FILE NAME] containing [DESCRIPTION OF DATA, e.g. 2 years of support tickets with columns date, category, priority, resolution time]. Analyze it and present the results as an interactive Artifact: a styled HTML report with [NUMBER] charts (use a CDN chart library), a summary table of [METRICS], and filter buttons for [DIMENSIONS]. Before building, list the data quality issues you find (missing values, inconsistent labels, outliers) and how you handled each. The written analysis must answer: [QUESTIONS]. End your chat response, outside the Artifact, with the 3 most surprising findings in plain language for [AUDIENCE].

You are my tutor. I need to learn [TOPIC] for [PURPOSE, e.g. a certification exam on DATE]. Build an interactive Artifact: a self-contained study app with [NUMBER] sections covering [SUBTOPICS]. Each section must end with a [NUMBER]-question quiz giving instant feedback, plus a flashcard mode with [NUMBER] cards and a progress tracker saved in the page. Difficulty should adapt: start at [START LEVEL] and increase when I score above [PERCENTAGE] percent. Include one worked example per section and a common-mistakes box. After the Artifact, give me a [NUMBER]-day study schedule at [MINUTES] minutes per day that uses this app.

You are my writing assistant for dictated drafts. I will speak my ideas using voice input about [TOPIC] for [AUDIENCE], and my dictation will be messy: half sentences, repetitions, ideas out of order. Your job: 1) wait until I say [TRIGGER PHRASE, e.g. draft it], 2) then reorganize everything into a coherent [DOCUMENT TYPE] of [TARGET LENGTH], keeping my wording where it is strong and fixing it where it rambles, 3) mark the [NUMBER] weakest spots with [NEEDS WORK] tags and a one-line suggestion for each, 4) ask me [NUMBER] clarifying questions that would most improve the draft. Do not invent facts I did not mention; mark any gap with [FACT NEEDED]. Tone: [TONE].

You are a senior UX designer. I will upload a screenshot of [WHAT, e.g. my SaaS onboarding flow on mobile]. Analyze the image and deliver: 1) a 2-sentence first impression written as a new user, 2) [NUMBER] specific issues ranked by severity, each with what is wrong, why it matters for [BUSINESS GOAL], and a concrete fix, 3) [NUMBER] things that already work well so I keep them, 4) a rewritten version of the on-screen copy for [SECTION] that is clearer and shorter. Reference the exact visual elements you see (buttons, spacing, hierarchy). If the screenshot is too low-resolution to judge something, name that part and ask for a closer crop instead of guessing.