More screenshots

    Works with the AI tools you already use

    CClaude CodeCCursorCCodex CLIGGitHub CopilotGGemini CLIVVS CodeWWindsurf+15 more

    Web UI Design-to-Code Prompt Architect

    1

    Transforms UI descriptions, screenshots, wireframes, Figma references, and existing frontend code into precise, implementation-ready prompts for AI coding agents. Defines page structure, component architecture, design tokens, responsive behavior, interaction states, accessibility, data boundaries, performance requirements, file plans, acceptance criteria, verification steps, and output expectations for React, Next.js, TypeScript, Tailwind CSS, shadcn/ui, HTML/CSS, Vue, Svelte, and component-based web applications.

    Secure checkout via Stripe

    0 installsSecurity scanned

    See it in action

    You say

    Project name: OrbitFlow Analytics Dashboard

    Page or component: Main SaaS analytics dashboard

    Input type: Desktop screenshot Mobile wireframe Existing Next.js repository Mixed input

    Primary user: Operations managers and customer-success leaders

    Primary user goal: Review critical account metrics, identify unusual changes, filter data by date, inspect recent activity, and export a report.

    Primary business goal: Help users recognize important account changes quickly and encourage regular dashboard usage.

    Target stack: Next.js React TypeScript Tailwind CSS shadcn/ui

    Framework: Next.js App Router

    Language: TypeScript with strict mode

    Styling system: Tailwind CSS using the existing project theme

    Component library: Installed shadcn/ui components Radix UI primitives already used by the project Lucide icons already installed

    Existing design system: Use existing color, typography, spacing, radius, and shadow tokens. Do not introduce a second component library.

    Existing project structure: App Router Routes inside app Shared components inside components Design-system components inside components/ui Feature-specific dashboard components inside features/dashboard Existing dashboard shell and sidebar must be reused

    Required route: app/dashboard/page.tsx

    Required components: DashboardHeader DateRangeControl MetricGrid MetricCard RevenueTrendPanel AccountHealthPanel RecentActivityTable DashboardEmptyState DashboardErrorState MobileDashboardNavigation

    Required interactions: Change date range Open account-health details Sort recent activity Filter activity by status Open activity record Start report export Open and close mobile navigation

    Required states: Initial loading Loaded No metrics No recent activity Filtered empty Dashboard error Table error Date range updating Export pending Export completed Export failed Permission restricted

    Data source: Reuse the existing typed dashboard service when available. Do not create new endpoints. Use typed mocked data only when a required service is missing. Keep the data boundary easy to replace.

    Responsive targets: 320-pixel mobile 390-pixel mobile Large mobile Tablet Laptop Desktop Wide desktop Short viewport height 200% browser zoom

    Accessibility target: WCAG 2.2 AA implementation review

    Browser support: Current stable versions of Chrome, Edge, Firefox, and Safari

    Localization: English initially German, Spanish, and French planned Allow at least 35% text expansion Do not concatenate sentence fragments Support locale-aware numbers and dates

    Performance constraints: Preserve server rendering where practical Avoid unnecessary client components Do not add large charting or animation dependencies Prevent layout shift Lazy-load secondary content where appropriate Respect reduced motion

    Allowed dependencies: Only dependencies already installed in the project

    Disallowed dependencies: New UI libraries New global state libraries New animation libraries Remote UI services

    Need output: A phased coding-agent prompt for Cursor or Claude Code Include repository inspection Component architecture File plan Responsive matrix Interaction states Accessibility Performance Acceptance criteria Verification Required final report

    Need variations: Conservative implementation that maximizes reuse More polished implementation using existing design tokens

    Special constraints: Do not rewrite the dashboard shell. Do not modify unrelated routes. Do not invent analytics values. Do not invent backend endpoints. Do not replace the existing table utilities. Do not turn the entire page into a client component. Do not remove working tests. Do not use page-level horizontal scrolling.

    Your agent does

    === PHASED DESIGN-TO-CODE IMPLEMENTATION PROMPT ===

    ROLE

    Act as a senior frontend engineer and design engineer specializing in Next.js App Router, React, TypeScript, Tailwind CSS, shadcn/ui, responsive SaaS dashboards, accessible component systems, and maintainable frontend architecture.

    OBJECTIVE

    Implement the OrbitFlow analytics dashboard using the supplied desktop screenshot, mobile wireframe, and existing Next.js repository.

    Preserve the screenshot's structural hierarchy and the wireframe's mobile information architecture while reusing the project's current dashboard shell, tokens, components, utilities, and data services.

    Do not treat the screenshot as proof of hidden interactions, exact dimensions, exact fonts, backend behavior, or complete responsive behavior.

    PRODUCT CONTEXT

    Primary users: Operations managers and customer-success leaders

    Primary user task: Review critical metrics, identify unusual changes, inspect recent activity, and export a report.

    Primary business goal: Make important account changes easy to recognize and support repeat dashboard use.

    Primary page hierarchy:

    1. Dashboard identity and date context
    2. Critical metrics
    3. Revenue trend
    4. Account-health issues
    5. Recent activity
    6. Export and secondary actions

    SOURCE INTERPRETATION

    Observed from the desktop screenshot:

    • Existing left sidebar
    • Dashboard header
    • Date-range control
    • Four metric cards
    • Large revenue chart
    • Smaller account-health panel
    • Recent activity table
    • Report export action
    • Neutral surfaces with one primary accent
    • Compact B2B SaaS visual density

    Observed from the mobile wireframe:

    • Top mobile navigation
    • Two-column metric grid when width permits
    • Stacked analytical panels
    • Recent activity represented as compact records
    • Date-range control near the page title
    • Export action inside an overflow menu

    Inferred:

    • Metric cards should be reusable.
    • Dashboard panels should share a common surface component.
    • Activity data requires sorting and status filtering.
    • Export is asynchronous.

    Requires verification:

    • Exact metric formulas
    • Exact chart library
    • Existing export service
    • Existing date-range API behavior
    • Exact permission model
    • Exact mobile breakpoint values

    TECHNOLOGY STACK

    Framework: Next.js App Router

    Language: TypeScript with strict mode

    Styling: Existing Tailwind CSS configuration

    Components: Installed shadcn/ui components and current project primitives

    Icons: Existing Lucide installation

    Data: Existing typed dashboard service when available

    Do not install additional dependencies unless the project cannot satisfy a required behavior and the user explicitly approves the dependency.

    PROJECT CONSTRAINTS

    • Reuse the current dashboard shell and sidebar.
    • Preserve existing routes.
    • Preserve existing analytics identifiers.
    • Preserve current table utilities.
    • Do not rewrite unrelated features.
    • Do not change the package manager.
    • Do not introduce a new global state library.
    • Do not convert the entire route into a client component.
    • Use client components only where interaction requires them.
    • Do not create new backend endpoints.
    • Do not hardcode private data.
    • Do not remove working tests.
    • Do not modify unrelated files.
    • Use existing project tokens before adding new values.
    • Avoid excessive Tailwind arbitrary values.
    • Avoid inline styles unless required by a verified chart API.
    • Avoid page-level horizontal scrolling.

    PHASE 1 — REPOSITORY INSPECTION

    Before implementing:

    1. Inspect app/dashboard/page.tsx.
    2. Identify the current dashboard shell and sidebar.
    3. Inspect components/ui.
    4. Inspect features/dashboard.
    5. Identify existing metric-card, panel, table, date-range, chart, status, loading, empty-state, and error-state components.
    6. Identify existing dashboard data services and types.
    7. Identify available verification scripts from package.json.
    8. Identify existing test conventions.
    9. Identify existing responsive and accessibility conventions.
    10. Report any conflict between this prompt and the current codebase.

    Do not begin a broad rewrite before completing this inspection.

    PHASE 2 — FILE PLAN

    Expected route: app/dashboard/page.tsx

    Potential feature files: features/dashboard/components/dashboard-header.tsx features/dashboard/components/date-range-control.tsx features/dashboard/components/metric-grid.tsx features/dashboard/components/metric-card.tsx features/dashboard/components/revenue-trend-panel.tsx features/dashboard/components/account-health-panel.tsx features/dashboard/components/recent-activity-table.tsx features/dashboard/components/recent-activity-mobile-list.tsx features/dashboard/components/dashboard-empty-state.tsx features/dashboard/components/dashboard-error-state.tsx features/dashboard/components/mobile-dashboard-navigation.tsx features/dashboard/types.ts features/dashboard/mock-data.ts

    Create only the files that are necessary and consistent with the repository.

    Reuse:

    • Existing dashboard shell
    • Existing sidebar
    • Existing UI primitives
    • Existing table utilities
    • Existing status badges
    • Existing loading skeletons
    • Existing error components
    • Existing data service
    • Existing chart library, if present

    Do not modify:

    • Unrelated routes
    • Authentication flows
    • Global application architecture
    • Backend services
    • Package manager configuration
    • Unrelated design-system primitives

    PHASE 3 — PAGE STRUCTURE

    Dashboard Page

    1. Existing dashboard shell
    2. Dashboard header
    3. Date-range and export actions
    4. Metric grid
    5. Primary analytics region
      • Revenue trend
      • Account health
    6. Recent activity
    7. Loading, empty, error, and permission states

    COMPONENT ARCHITECTURE

    DashboardPage ├── ExistingDashboardShell ├── DashboardHeader │ ├── PageTitle │ ├── DateRangeControl │ └── DashboardActions ├── MetricGrid │ └── MetricCard ├── DashboardAnalyticsGrid │ ├── RevenueTrendPanel │ └── AccountHealthPanel ├── RecentActivitySection │ ├── ActivityToolbar │ ├── RecentActivityTable │ └── RecentActivityMobileList ├── DashboardEmptyState ├── DashboardErrorState └── MobileDashboardNavigation

    COMPONENT RESPONSIBILITIES

    DashboardHeader:

    • Display page title and contextual supporting text.
    • Contain date-range and export actions.
    • Transform actions into an accessible mobile arrangement.
    • Preserve semantic heading order.

    MetricCard:

    • Accept typed label, value, trend, trend direction, supporting context, loading state, and optional action.
    • Use one shared structure for all metrics.
    • Communicate trend direction using text or icons in addition to color.
    • Support long translated values and labels.

    RevenueTrendPanel:

    • Display the existing revenue visualization.
    • Include a textual summary for accessibility.
    • Preserve loading, empty, and error states.
    • Avoid continuous animation.
    • Respect reduced motion.

    AccountHealthPanel:

    • Prioritize actionable issues over decorative metrics.
    • Show severity, label, count, and verified action.
    • Support an empty state when no issues exist.

    RecentActivityTable:

    • Reuse existing table utilities.
    • Support sorting.
    • Support status filtering.
    • Use semantic headers.
    • Include accessible row actions.
    • Provide populated, loading, filtered-empty, error, and permission states.

    RecentActivityMobileList:

    • Display essential record information without page-level horizontal scrolling.
    • Preserve the same data and actions as the desktop table.
    • Do not make swipe the only interaction method.

    VISUAL SYSTEM

    Use existing project tokens.

    Visual direction:

    • Premium B2B SaaS
    • Calm neutral surfaces
    • One primary accent
    • Strong information hierarchy
    • Limited shadow usage
    • Compact but readable data presentation
    • Clear priority for account-health issues

    Typography roles:

    • Page title
    • Section title
    • Metric value
    • Metric label
    • Supporting text
    • Table label
    • Metadata
    • Button label

    Spacing:

    • Use the existing spacing scale.
    • Keep related labels and values close.
    • Use larger separation between major dashboard regions.
    • Keep mobile gutters consistent.
    • Avoid overpadding metric cards.

    Cards and panels:

    • Reuse the existing surface component.
    • Use one consistent border and radius treatment.
    • Do not give every panel the same visual emphasis.
    • Make the main trend panel stronger than secondary informational panels.
    • Make critical account-health states visually clear without relying on color alone.

    RESPONSIVE MATRIX

    Wide Desktop:

    • Persistent sidebar
    • Four metric cards in one row where supported
    • Revenue panel wider than account-health panel
    • Full activity table

    Desktop:

    • Persistent sidebar
    • Four-column or two-column metric grid based on available content width
    • Two-column analytics region
    • Full activity table

    Tablet:

    • Collapsible sidebar or existing tablet navigation
    • Two-column metric grid
    • Single-column analytical panels when required
    • Activity table may retain priority columns

    Mobile:

    • Accessible top navigation
    • One or two metric columns depending on available width
    • Full-width date control
    • Export action inside an accessible menu where needed
    • Stacked analytical panels
    • Card-based recent activity
    • No page-level horizontal scrolling

    Small Mobile:

    • Single-column metric cards
    • Wrapped labels
    • Full-width primary controls
    • No clipped values
    • No overlapping actions

    Short Viewport:

    • Sticky controls must not cover content or focus.
    • Mobile navigation must remain operable with browser chrome and the virtual keyboard.

    200% Zoom:

    • Content must reflow.
    • Controls must remain reachable.
    • Tables must provide an accessible alternative.
    • Text must not overlap or become clipped.

    INTERACTION STATES

    Date Range:

    • default
    • focus-visible
    • open
    • selected
    • updating
    • error
    • disabled

    Export:

    • idle
    • pending
    • completed
    • failed
    • permission restricted

    Dashboard Data:

    • loading
    • loaded
    • empty
    • partial
    • error
    • stale where supported

    Activity:

    • loading
    • populated
    • no activity
    • filtered empty
    • sorting
    • filtering
    • table error
    • permission restricted

    ACCESSIBILITY REQUIREMENTS

    • Use semantic landmarks.
    • Preserve one clear H1.
    • Maintain logical heading order.
    • Provide accessible names for every icon-only control.
    • Preserve visible focus.
    • Support keyboard operation for date controls, menus, sorting, filters, and row actions.
    • Associate control labels and descriptions.
    • Use proper table headers.
    • Communicate sorting state programmatically.
    • Communicate trend direction beyond color.
    • Provide a textual chart summary.
    • Manage focus when the mobile navigation or action menu opens.
    • Return focus to the trigger when overlays close.
    • Support reduced motion.
    • Ensure error states are visible and announced appropriately.
    • Support zoom and responsive reflow.
    • Keep touch targets usable.
    • Do not claim accessibility conformance without testing.

    DATA BOUNDARIES

    Use the existing typed dashboard service when available.

    Do not create new endpoints.

    When a required service is missing:

    • use typed mock data
    • isolate mock data from presentation components
    • document the replacement boundary
    • do not present mocked values as live production data

    Required models may include:

    • DashboardMetric
    • MetricTrend
    • RevenueDataPoint
    • AccountHealthIssue
    • ActivityRecord
    • ActivityStatus
    • DateRange
    • ExportState

    PERFORMANCE REQUIREMENTS

    • Preserve server rendering where practical.
    • Keep interactive controls in focused client components.
    • Avoid making the entire page a client component.
    • Reuse the installed chart library.
    • Do not add a new charting dependency.
    • Reserve chart dimensions to reduce layout shift.
    • Lazy-load secondary analytical content only when it improves performance without harming usability.
    • Avoid continuous animation.
    • Respect reduced motion.
    • Prevent unnecessary rerenders.
    • Avoid duplicated data transformation.
    • Keep component boundaries focused.

    IMPLEMENTATION SEQUENCE

    1. Inspect repository conventions.
    2. Confirm the route and reusable components.
    3. Confirm existing data services and types.
    4. Create or refine the dashboard data model.
    5. Implement the structural page layout.
    6. Build or reuse metric cards and panels.
    7. Implement desktop activity table.
    8. Implement mobile activity presentation.
    9. Add loading, empty, error, partial, and permission states.
    10. Add responsive behavior.
    11. Add accessibility behavior.
    12. Add or update tests where supported.
    13. Run available verification.
    14. Summarize implementation and remaining assumptions.

    ACCEPTANCE CRITERIA

    Dashboard Structure:

    Given the dashboard loads on desktop, when the page is displayed, then the page title, date context, critical metrics, revenue trend, account health, and recent activity appear in a clear priority order.

    Component Reuse:

    Given shared dashboard components already exist, when implementation begins, then the solution reuses or extends those components rather than duplicating equivalent structures.

    Mobile Layout:

    Given the page is displayed at 390 pixels wide, when the dashboard is loaded, then all primary metrics, actions, analytical panels, and activity records remain accessible without page-level horizontal scrolling.

    Date Update:

    Given the user changes the date range, when new data is loading, then the control and affected dashboard regions communicate the updating state without removing the current context prematurely.

    Export:

    Given the user starts an export, when processing begins, then the export action communicates the pending state, prevents unsafe duplicate submission, and reports completion or failure.

    Activity Empty State:

    Given no records match the active status filter, when filtering completes, then the interface displays a filtered-empty state with a clear method to remove or change the filter.

    Keyboard Navigation:

    Given a keyboard user opens the mobile navigation, when the menu appears, then focus enters the menu, remains managed while open, and returns to the trigger when closed.

    Zoom:

    Given the page is viewed at 200% zoom, when content reflows, then text, controls, metrics, panels, and activity records remain readable and operable without overlap.

    VERIFICATION

    Run only scripts supported by the project.

    Verify where available:

    • TypeScript checking
    • Linting
    • Production build
    • Relevant component tests
    • Relevant end-to-end tests
    • Responsive inspection
    • Keyboard navigation
    • Visible focus
    • Mobile menu focus behavior
    • Loading states
    • Empty states
    • Error states
    • Export states
    • 320-pixel mobile
    • 390-pixel mobile
    • Tablet
    • Desktop
    • 200% zoom
    • Reduced motion

    REQUIRED FINAL RESPONSE

    Return:

    1. Implementation summary
    2. Files created
    3. Files modified
    4. Existing components reused
    5. Component architecture decisions
    6. Responsive behavior implemented
    7. Accessibility behavior implemented
    8. Data assumptions
    9. Verification completed
    10. Verification not completed
    11. Remaining risks
    12. Open questions

    PROHIBITED BEHAVIOR

    • Do not rewrite the dashboard shell.
    • Do not modify unrelated routes.
    • Do not install unnecessary dependencies.
    • Do not create new backend endpoints.
    • Do not invent analytics values.
    • Do not remove working tests.
    • Do not replace existing table utilities.
    • Do not convert the complete route into a client component.
    • Do not use one monolithic dashboard component.
    • Do not use inaccessible icon-only controls.
    • Do not rely on color alone.
    • Do not introduce page-level horizontal scrolling.
    • Do not hardcode secrets.
    • Do not expose private data.
    • Do not claim verification that was not completed.

    UNCERTAINTY HANDLING

    When a detail is missing:

    • inspect the existing project first
    • choose the smallest implementation consistent with project conventions
    • label the assumption
    • avoid inventing backend behavior
    • preserve a replaceable data boundary
    • document the remaining verification requirement

    FINAL INTEGRITY NOTE

    All backend behavior, authentication, authorization, data persistence, analytics, privacy, security, payment, and production assumptions must be verified before implementation.

    Generated code must be reviewed, tested, and adapted to the actual project architecture before production use.

    The prompt architecture is designed to improve code-generation consistency, visual fidelity, maintainability, accessibility readiness, and implementation clarity. It does not guarantee defect-free, secure, compliant, production-ready, or pixel-identical output.

    What you get

    Convert screenshots into structured React component hierarchies.Generate responsive implementation rules for complex UI layouts.Refactor monolithic frontend files into clean, modular components.Specify accessibility and interaction states for AI-generated code.Bridge the gap between design handoffs and implementation-ready prompts.

    About this skill

    Web UI Design-to-Code Prompt Architect helps frontend developers, Cursor users, Claude Code users, Replit users, Codex CLI users, no-code builders, startup founders, product designers, and development teams convert incomplete interface ideas into precise prompts that AI coding agents can implement consistently.

    The skill accepts natural-language UI descriptions, screenshots, wireframes, Figma exports, design references, product requirements, existing frontend code, component inventories, visual redesign requests, responsive specifications, and accessibility requirements.

    It converts those inputs into structured coding instructions for React, Next.js, TypeScript, JavaScript, Tailwind CSS, shadcn/ui, Radix UI, HTML/CSS, Vue, Nuxt, Svelte, SvelteKit, and other component-based frontend architectures.

    Instead of producing weak prompts such as “build a modern SaaS dashboard,” the skill defines exactly what the coding agent should build, how the page should be structured, which components should be reusable, how the design should behave across breakpoints, which interaction states are required, what data assumptions are safe, which dependencies are allowed, and how implementation quality should be verified.

    The agent begins by identifying the project context, page purpose, target user, primary task, business objective, source type, technology stack, existing codebase conventions, implementation constraints, and required output format.

    For screenshot-based work, it separates visible evidence from inferred behavior. It identifies observed layout, hierarchy, components, typography roles, spacing, color direction, imagery, cards, controls, and responsive clues without pretending that a screenshot reveals hidden interactions, exact breakpoints, exact fonts, backend logic, or data behavior.

    For wireframes, it preserves the information architecture and user flow while creating a coherent visual system, component hierarchy, responsive strategy, interaction-state model, and accessibility specification.

    For written UI descriptions, it converts vague adjectives such as modern, elegant, premium, clean, or responsive into measurable implementation rules covering layout, spacing, type hierarchy, color roles, borders, shadows, component density, CTA prominence, and breakpoint behavior.

    For existing frontend code, the skill analyzes framework conventions, routes, component architecture, styling methods, state management, dependencies, design tokens, accessibility patterns, reusable utilities, tests, and regression risks. The resulting prompt instructs the coding agent to preserve working behavior, reuse existing utilities, avoid unnecessary rewrites, modify the smallest coherent surface, and document every changed file.

    The skill creates complete prompt architecture using twenty implementation layers:

    1. Coding-agent role
    2. Objective
    3. Product and user context
    4. Source interpretation
    5. Technology stack
    6. Project constraints
    7. Page and route structure
    8. Component architecture
    9. Visual system
    10. Responsive behavior
    11. Interaction states
    12. Accessibility
    13. Data assumptions
    14. Performance requirements
    15. Implementation sequence
    16. Acceptance criteria
    17. Verification
    18. Required response format
    19. Prohibited behavior
    20. Uncertainty handling

    Page structures can be defined region by region. Each region may include purpose, content, layout, hierarchy, primary action, component dependencies, responsive behavior, interaction states, and accessibility requirements.

    Component architecture can include page shells, navigation, headers, sidebars, cards, forms, tables, filters, dialogs, drawers, alerts, charts, empty states, loading states, error states, success states, and reusable design-system primitives.

    The skill can produce explicit component trees and specifications covering responsibility, props, local state, shared state, variants, events, accessibility, responsive transformations, and reuse.

    Visual-system instructions can define typography roles, color roles, page widths, grid behavior, gutters, spacing scales, component gaps, card padding, border radii, borders, shadows, iconography, imagery, motion, focus treatment, and semantic status colors.

    It avoids arbitrary hardcoded values when an existing design system is available and instructs coding agents to reuse project tokens, Tailwind theme values, CSS custom properties, existing component variants, and established visual conventions.

    Responsive behavior is specified explicitly for wide desktop, desktop, tablet, mobile, small mobile, short viewport heights, and zoomed layouts where relevant.

    The prompt can define which elements stack, reorder, collapse, become drawers, become full width, remain sticky, transform into cards, become horizontally contained, or disappear because they are genuinely nonessential.

    It prohibits page-level horizontal scrolling, clipped labels, overlapping controls, inaccessible touch targets, hidden primary actions, hover-only functionality, gesture-only interactions, and sticky elements that cover focused content.

    Interaction-state architecture can include default, hover, focus-visible, active, selected, disabled, loading, empty, filtered empty, error, success, warning, offline, partial success, permission restricted, stale, updating, retrying, cancelled, and completed states.

    Form prompts can define labels, descriptions, validation, required and optional conventions, field errors, form-level errors, submission states, server errors, success states, keyboard behavior, and focus recovery.

    Data specifications distinguish static content, mocked content, typed props, existing services, server-loaded data, client-loaded data, and verified API integrations. The agent is instructed not to invent endpoints, schemas, authentication, authorization, payment logic, or persistence behavior.

    Accessibility requirements can cover semantic landmarks, heading order, accessible names, form labels, descriptions, error relationships, keyboard support, focus-visible treatment, dialog focus management, reduced motion, non-color state communication, zoom, reflow, touch targets, image alternative text, table semantics, and live announcements.

    Performance guidance can cover server and client component boundaries, image optimization, lazy loading, layout stability, bundle growth, unnecessary dependencies, expensive visual effects, animation, rerenders, large lists, code splitting, and secondary-content loading.

    The skill creates file-by-file implementation plans that identify files to create, files to modify, existing components to reuse, files that should not be changed, and possible shared primitives.

    It can produce one-shot prompts for limited tasks or phased implementation prompts for complex applications. Phases may include repository inspection, design foundations, structural implementation, component creation, states, responsive behavior, accessibility, testing, and final polish.

    Acceptance criteria are written as measurable requirements or Given/When/Then scenarios. They can cover responsive behavior, component reuse, visual hierarchy, accessibility, loading, errors, form behavior, keyboard navigation, dialogs, mobile menus, data states, and regression protection.

    Verification instructions can require available type checking, linting, builds, unit tests, component tests, end-to-end tests, accessibility checks, responsive inspection, keyboard review, focus review, visual inspection, and changed-file reporting.

    The coding agent is instructed to report what it implemented, which files were created or changed, which components were reused, which checks were completed, which checks could not be completed, which assumptions remain, and which risks still require review.

    The skill supports several specialized output modes:

    Complete design-to-code implementation prompt Screenshot-to-code prompt Wireframe-to-code prompt Description-to-code prompt Existing-code refactor prompt Responsive redesign prompt Accessibility-remediation prompt Component-extraction prompt Design-system implementation prompt Phased coding-agent plan Frontend prompt audit Structured JSON implementation specification

    The prompt-audit mode evaluates objective clarity, source interpretation, stack definition, project constraints, page structure, component architecture, visual specificity, responsive behavior, state coverage, accessibility, data assumptions, performance, code quality, implementation sequence, acceptance criteria, verification, prohibited behavior, and uncertainty handling.

    The skill helps prevent common AI-generated frontend failures such as monolithic components, duplicated styles, arbitrary colors, inconsistent spacing, weak semantic HTML, inaccessible controls, missing focus states, incomplete error handling, unsupported dependencies, invented APIs, poor TypeScript types, incorrect responsive behavior, excessive client rendering, unnecessary animation, missing loading states, and vague completion criteria.

    It prioritizes visual fidelity, implementation clarity, component reuse, accessibility readiness, responsive quality, maintainability, controlled dependencies, safe assumptions, realistic verification, and compatibility with the existing project architecture.

    The core commercial promise is: convert vague UI ideas into precise coding-agent prompts that produce more consistent, maintainable, responsive, accessible, and implementation-ready frontend results.

    How to install

    Drop the file into your AI Agent. Works with Claude, Cursor, ChatGPT, and 20+ more.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    30-day refund

    Not a fit? Get your money back

    Trust & safety

    Security scanned

    Verified clean today

    Listedtoday

    Frequently Asked Questions