agentleFS
KilimiaoSixKilimiaoSix/agent-skills/skills/frontend-design-codex/SKILL.md

frontend-design-codex

Skill1 starsChanged 4 months agoMITView on GitHub

Create distinctive, production-grade frontend interfaces in Codex. Use when the user asks to build, redesign, restyle, modernize, or visually polish web components, pages, dashboards, apps, prototypes, games, hero sections, or existing UI, especially when they want non-generic AI aesthetics, strong typography, color, motion, layout, responsive behavior, and browser-verified implementation.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

39 rules. About 0 are common, 38 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈38

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Commit to one distinctive visual direction before coding.

    line 15
  2. Distinctive:

    Verify in a browser on desktop and mobile sizes.

    line 18
  3. Distinctive:

    Before coding, write a compact design brief in the working notes:

    line 24
  4. Distinctive:

    If the request is a full page, app, dashboard, or visually led redesign, use Image Gen for concepting unless the user opts out, an existing …

    line 32
  5. Distinctive:

    Commit to a direction that fits the domain.

    line 36
  6. Distinctive:

    Avoid generic AI frontend habits.

    line 62
  7. Distinctive:

    Do not default to purple-blue gradients on white backgrounds.

    line 64
  8. Distinctive:

    Do not default to Inter, Roboto, Arial, or system fonts when a distinctive but readable font pairing is appropriate.

    line 65
  9. Distinctive:

    Do not use repeated card grids as the main structure unless cards are the actual interaction model.

    line 66
  10. Distinctive:

    Do not add hero eyebrow labels, pill soup, fake badges, fake metrics, or logo clouds unless the brief or reference requires them.

    line 67
  11. Distinctive:

    Do not make every section a centered headline plus three cards.

    line 68
  12. Distinctive:

    Do not make dashboard data into marketing cards when tables, timelines, maps, charts, or inspectors are the real workflow.

    line 69
  13. Distinctive:

    Do not add decorative glows, blobs, or gradients that do not clarify hierarchy or brand.

    line 70
  14. Distinctive:

    Do not use stock-like generic imagery that could fit any product.

    line 71
  15. Distinctive:

    Do not let browser-default control typography appear in buttons, tabs, inputs, sidebars, toolbars, table cells, or status bars.

    line 72
  16. Distinctive:

    Do not ship inert controls in app UI.

    line 73
  17. Distinctive:

    Do not use tiny low-contrast text to make the design look polished while hurting readability.

    line 74
  18. Distinctive:

    Do not use the same type scale, radius, shadows, and palette across unrelated prompts.

    line 75
  19. Distinctive:

    Do not hide weak hierarchy behind excessive animation.

    line 76
  20. Distinctive:

    Do not use a static screenshot as app UI.

    line 77
  21. Distinctive:

    Do not change a true white reference background into cream or gray unless the concept calls for it.

    line 78
  22. Distinctive:

    Before implementation, define the local system in code or notes:

    line 82
  23. Distinctive:

    Follow the repo's framework, routing, component, state, styling, lint, and accessibility patterns.

    line 96
  24. Distinctive:

    Prefer existing icon libraries when they match the direction.

    line 100
  25. Distinctive:

    Use CSS variables or theme tokens for all recurring colors, spacing, and typography.

    line 101
  26. Distinctive:

    Keep text readable at common desktop and mobile widths.

    line 102
  27. Very specific:

    Add accessibility basics: semantic landmarks, button labels, focus states, sufficient contrast, and prefers-reduced-motion.

    line 103
  28. Distinctive:

    Keep dependencies local to the project.

    line 104
  29. Distinctive:

    Check at least one desktop and one mobile viewport.

    line 108
  30. Distinctive:

    Use motion to clarify hierarchy, state, and tangibility.

    line 119
  31. Distinctive:

    Prefer 2-3 deliberate motions over many scattered effects.

    line 121
  32. Distinctive:

    Avoid motion that delays task completion, hides content, or causes layout shift.

    line 123
  33. Distinctive:

    Verify at least one animated or interactive state in the browser.

    line 125
  34. Distinctive:

    Run the app and verify the visible product, not just the build.

    line 129
  35. Distinctive:

    Start the local dev server when needed and record the URL.

    line 131
  36. Distinctive:

    Prefer the Codex Browser plugin for local targets.

    line 132
  37. Distinctive:

    Check desktop and mobile widths.

    line 133
  38. Distinctive:

    Keep fixing if there is clipped content, unreadable text, inert controls, placeholder visuals, stale debug UI, mobile overflow, or a …

    line 137
  39. Distinctive:

    Do not claim visual success without browser evidence when the task involves a runnable UI.

    line 151

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Frontend Design Codex

Use this skill to turn frontend requests into working, production-oriented UI with a clear aesthetic point of view. The goal is not to decorate a default layout. The goal is to design and implement an interface that feels specific to the product, audience, and interaction model.

This skill adapts the public Claude frontend-design workflow to Codex. It uses Codex-native capabilities: local file edits, Image Gen when concepting is useful, Browser or Playwright verification, existing project design systems, and mechanical handoff evidence.

Quick Workflow

  1. Understand the product, audience, workflow, constraints, and success state.
  2. Commit to one distinctive visual direction before coding.
  3. Build a small design system: tokens, typography, spacing, components, motion, and asset treatment.
  4. Implement the real usable surface in the repo's existing stack.
  5. Verify in a browser on desktop and mobile sizes.
  6. Iterate until the visible UI matches the chosen direction and has no obvious layout, responsiveness, typography, or interaction defects.
  7. Final-answer with changed files, verification commands, browser URL, and remaining intentional deviations.

Design Pre-Analysis

Before coding, write a compact design brief in the working notes:

  • Purpose: the problem the interface solves and the primary user.
  • Context: landing page, operational app, dashboard, editor, commerce flow, game, or component.
  • Tone: choose one precise direction, such as editorial utility, industrial control room, playful toy-like, luxury restraint, brutalist raw, retro-futuristic, organic/natural, art deco, dense analyst workspace, or calm enterprise tool.
  • Differentiator: the one visual or interaction idea a user should remember.
  • Constraints: framework, existing design system, accessibility, performance, content, data shape, and responsive requirements.

If the request is a full page, app, dashboard, or visually led redesign, use Image Gen for concepting unless the user opts out, an existing design system already dictates the visual answer, or the task is a narrow UI fix.

Visual Direction Rules

Commit to a direction that fits the domain. Do not mix several moods because they sound impressive.

  • SaaS, CRM, admin, analytics, and operational tools should be dense, calm, legible, and workflow-first.
  • Consumer, brand, venue, editorial, and product pages should have a strong first-viewport signal and real visual assets.
  • Games and playful demos can be more expressive, illustrated, and animated, but core rules and controls must work.
  • Existing products should preserve recognizable information architecture before adding visual polish.

Style Range Matrix

This skill is not a single visual style. If the user does not specify a style and the task is visually led, consider three meaningfully different directions before choosing one. Pick the direction that best fits the product, audience, and workflow, then make the implementation coherent around that choice.

Useful direction families:

  • Editorial utility: strong type hierarchy, restrained color, print-like rhythm, useful for publishing, research, documentation, and serious brand sites.
  • Industrial control room: dense rows, hard dividers, status color, mono details, useful for operations, monitoring, logistics, and infrastructure tools.
  • Playful toy-like: chunky controls, kinetic motion, friendly color, useful for education, kids, games, and creative demos.
  • Luxury restraint: large negative space, tactile imagery, refined typography, useful for fashion, fragrance, hospitality, and premium products.
  • Brutalist raw: exposed grids, sharp contrast, direct labeling, useful for experimental portfolios, events, and opinionated cultural work.
  • Retro-futuristic: period-specific type, geometric panels, glow used sparingly, useful for music, games, hardware, and speculative products.
  • Organic/natural: soft asymmetry, material texture, botanical or earthy color, useful for wellness, food, environment, and handmade products.
  • Art deco: geometric ornament, high contrast, vertical rhythm, useful for venues, editorial campaigns, and premium event pages.
  • Dense analyst workspace: compact tables, filters, inspectors, charts, and keyboard-friendly flows for professional tools.
  • Clinical calm: high legibility, neutral surfaces, trustworthy states, useful for healthcare, finance, security, and compliance.

Anti-Generic UI Rules

Avoid generic AI frontend habits. Apply at least 10 of these checks before handoff:

  1. Do not default to purple-blue gradients on white backgrounds.
  2. Do not default to Inter, Roboto, Arial, or system fonts when a distinctive but readable font pairing is appropriate.
  3. Do not use repeated card grids as the main structure unless cards are the actual interaction model.
  4. Do not add hero eyebrow labels, pill soup, fake badges, fake metrics, or logo clouds unless the brief or reference requires them.
  5. Do not make every section a centered headline plus three cards.
  6. Do not make dashboard data into marketing cards when tables, timelines, maps, charts, or inspectors are the real workflow.
  7. Do not add decorative glows, blobs, or gradients that do not clarify hierarchy or brand.
  8. Do not use stock-like generic imagery that could fit any product.
  9. Do not let browser-default control typography appear in buttons, tabs, inputs, sidebars, toolbars, table cells, or status bars.
  10. Do not ship inert controls in app UI; tabs, filters, selections, drawers, and primary actions need visible state.
  11. Do not use tiny low-contrast text to make the design look polished while hurting readability.
  12. Do not use the same type scale, radius, shadows, and palette across unrelated prompts.
  13. Do not hide weak hierarchy behind excessive animation.
  14. Do not use a static screenshot as app UI; keep navigation, text, controls, forms, and tables code-native.
  15. Do not change a true white reference background into cream or gray unless the concept calls for it.

Design System Extraction

Before implementation, define the local system in code or notes:

  • Colors: background, surface, text, muted text, border, accent, semantic states, shadows.
  • Typography: display, heading, body, label, caption, control text, weights, line heights, and fallback fonts.
  • Spacing: page gutters, section rhythm, grid columns, panel padding, control heights, row density.
  • Shape: radii, borders, dividers, elevation, media frame treatment.
  • Components: buttons, links, tabs, filters, rows, cards only where necessary, tables, charts, forms, modals, empty states.
  • Motion: page entrance, hover/focus, selection, reveal, loading, and reduced-motion behavior.
  • Assets: real or generated images, product renders, icons, logos, textures, diagrams, sprites, and their crop/aspect rules.

Repeated UI must use shared tokens or reusable components. Differences should be named variants, not one-off copied styling.

Implementation Rules

  • Follow the repo's framework, routing, component, state, styling, lint, and accessibility patterns.
  • For a new complex frontend without an existing stack, default to React + Vite. For a simple static proof, HTML/CSS/JS is acceptable.
  • Build the real usable surface first. Do not wrap a future app in a marketing shell unless the user asked for a landing page.
  • Preserve user-provided copy, required data, navigation labels, CTA labels, states, and workflows.
  • Prefer existing icon libraries when they match the direction; otherwise write small, clean SVGs with consistent optical size.
  • Use CSS variables or theme tokens for all recurring colors, spacing, and typography.
  • Keep text readable at common desktop and mobile widths. Fix wrapping, clipping, and overlap before final handoff.
  • Add accessibility basics: semantic landmarks, button labels, focus states, sufficient contrast, and prefers-reduced-motion.
  • Keep dependencies local to the project. Do not add a package unless it materially improves the result and the user or repo allows it.

Responsive Rules

Check at least one desktop and one mobile viewport.

  • First viewport must fit without clipped primary content.
  • Navigation must collapse or simplify cleanly.
  • Tables, inspectors, editors, and dashboards must preserve scanning value; do not turn every dense surface into stacked cards automatically.
  • Buttons and controls need stable sizes and tap targets.
  • Hero or main workspace text must not overlap imagery, charts, or controls.
  • Fixed or sticky headers count against viewport height.

Motion Rules

Use motion to clarify hierarchy, state, and tangibility.

  • Prefer 2-3 deliberate motions over many scattered effects.
  • Good defaults: staggered entrance, state transition, hover/focus affordance, scroll-linked reveal, or active selection.
  • Avoid motion that delays task completion, hides content, or causes layout shift.
  • Respect prefers-reduced-motion.
  • Verify at least one animated or interactive state in the browser.

Browser Verification

Run the app and verify the visible product, not just the build.

  1. Start the local dev server when needed and record the URL.
  2. Prefer the Codex Browser plugin for local targets. Use Playwright when Browser is unavailable or when scripted screenshots are more reliable.
  3. Check desktop and mobile widths.
  4. Click through the core workflow: navigation, filter, tab, form, playback, selection, game control, or primary CTA.
  5. Inspect typography, color, spacing, responsive behavior, icon treatment, asset loading, focus states, and motion.
  6. Capture a screenshot when practical and compare it against the concept, reference, or written visual direction.
  7. Keep fixing if there is clipped content, unreadable text, inert controls, placeholder visuals, stale debug UI, mobile overflow, or a generic template-like result.

Handoff

Final response should be concise and evidence-based:

  • Files changed.
  • Browser URL or static file path.
  • Build/test/lint commands run and their result.
  • Desktop and mobile verification method.
  • Core interaction path verified.
  • At least five visual checks performed.
  • Intentional deviations or remaining limitations.

Do not claim visual success without browser evidence when the task involves a runnable UI.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.