Agentic prompts, reviewed and ready to paste.
Copy a prompt, paste it into your coding assistant, get the work done. You see what it touches before you run it.
- Reviewed by a human
- Declares what it touches
- Comes with an undo path
- Copy a prompt
- Paste it into your coding assistant›
- It runs. You see what it touches.→Reading your project · building a plan!Touches: 3 files, 1 external tool · asks before writing✓Done · every change reversible
Featured
Hand-picked this week.
A senior web designer inside Claude Code. One paste sets it up; then you ask for a page in plain words and get a modern site with an immersive experience: a real design direction, real components from 21st.dev, and motion that tells the story.
Claude as co-architect of an AI-ready design system: tokens, contracts, components, Storybook and Figma generated from one repository, with you approving every gate. Install it as a Claude Code plugin or run it as a prompt.
Sets up a UX benchmark analyst in your assistant once (browser, PowerPoint, skills, your sector and reference sites). Then "benchmark the size guide" gives you ~20 best-in-sector sites, 3–4 screens per device showing the flow, each annotated with sourced best practice, in a modern PowerPoint.
Latest additions
The newest prompts in the catalogue.
A senior UX auditor inside Claude Code. One paste sets it up; then you name a site and a flow and it walks it in your Chrome on desktop and mobile, captures evidence step by step and hands you a report: findings rated by severity, the heuristic or WCAG criterion behind each one, and what to do about it.
Give it the URL of a site and get its design DNA as a usable design system: colour ramps, type roles and scale, spacing, radii, shadows, motion, a component inventory, delivered as W3C tokens, CSS variables, Tailwind and shadcn themes, Figma variables and a brand book. Values, never code.
A product video director inside Claude Code. One paste sets it up; then "make a 20-second promo of the new size guide for TikTok" gives you renders built to be watched to the end and to convert: a hook in the first 3 seconds in three variants to test, the product as the answer, one call to action, captions for sound-off viewing, your brand tokens, safe zones respected. Written in code with Remotion, so every version is reproducible.
Sets up a Jira agent in your assistant once (Jira and Figma via MCP, saved skills, ticket style). Then you say what to open, from Figma or in your words, check the preview, approve, and everything is created in Jira automatically.
Your GA4 analyst on request: custom reports, ecommerce KPI questions, year-over-year analysis, funnels, tracking audits and before/after readouts of a release. With no MCP available it works through the Claude in Chrome extension.
Keeps your GTM container clean and builds what you ask: tags, triggers and variables from plain language, prepared in a workspace, tested in Preview, published only by you.
Browse by category
Every prompt lives in exactly one category.
Web design & development
Design systems, modern websites, storefront development5 promptsUX research
Benchmarks, audits, heuristics2 promptsDelivery & PM
Jira, ClickUp, handoff, meeting notes2 promptsAnalytics
GA4, Contentsquare, KPI diagnosis2 promptsTagging & experimentation
GTM, data layer, A/B tests2 promptsTesting & QA
Playwright, BrowserStack, regressions1 promptThree steps. Your coding assistant does the typing.
- 01Pick a promptRead what it does, what it touches and how to undo it. Check reviews from people who ran it.
- 02Copy and pasteOne click copies the prompt. Paste it into Claude Code, Codex, Cursor or any agentic assistant.
- 03Follow alongThe assistant asks before risky steps. When it is done, rate the prompt and help the next person.
Built a prompt that saves people hours?
Submit it. Reviewers check every prompt for safety and reversibility. You keep the credit.