frontend-design-system
Skill4 starsChanged 37 days agoMITView on GitHub
Apply the user's theme-first frontend design system to UI work. Use whenever a task creates, edits, restyles, reviews, or diagnoses HTML/CSS, React/Vue UI, dashboards, workspaces, visual reports, components, color palettes, typography, spacing, responsive layout, or screenshot-based visual references. Use for both small visual fixes and full redesigns; preserve existing project language for small changes and perform theme extraction before new-page or redesign work.
9 other projects use a version of this.
Adopted by
Most recently taken up by
- AntonioMellies/claude-code-skillstheir own version, first seen Nov 1, 2026
- hushh-labs/hushh-researchtheir own version, first seen Nov 1, 2026
- ChainChase/agent-toolboxtheir own version, first seen Oct 15, 2026
- gongyijie85/dsh-ecctheir own version, first seen Oct 13, 2026
- mnzralee/claude-multi-agent-architecturetheir own version, first seen Oct 3, 2026
- polyant-ai/polyanttheir own version, first seen Sep 11, 2026
Used with
Projects with it also use verification.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
OpenAISkillverification
Full-story verification — infers what the user is building, then verifies the complete flow end-to-end: browser → API → data → response. Triggers on dev server start and 'why isn't this working' signals.
Also in 3 projects that use this
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- oikon48/cc-frontend-skills★ 69same name
- hushh-labs/hushh-research★ 40same name
- polyant-ai/polyant★ 10same name
- gongyijie85/dsh-ecc★ 8same name
- mnzralee/claude-multi-agent-architecture★ 8same name
- sefaertunc/Worclaude★ 4same name
- AntonioMellies/claude-code-skills★ 2same name
- sefaertunc/anthropic-watch★ 1same name
- ChainChase/agent-toolbox★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository29 other files it gives its agents.
- analysis-report-orchestratorSkill
- chatlab-analyzerSkill
- document-to-markdownSkill
- ecom-image-reverse-sop-shareSkill
- ecom-main-image-diagnosisSkill
- ecom-market-insight-tableSkill
- ecommerce-competitor-analyzerSkill
- ecommerce-data-analysis-workflowSkill
- ecommerce-review-analysisSkill
- ecom-one-photo-pageSkill
- ecom-page-plannerSkill
- ecom-product-assets-collectorSkill
- ecom-ranking-main-image-collectorSkill
- ecom-video-workflowSkill
- external-skill-adapterSkill
- feishu-project-weekly-reportSkill
- full-spectrum-distillerSkill
- github-repo-dissectorSkill
- kolsprite-tiktok-opsSkill
- manim-hyperframes-videoSkill
- rpa-flow-architectSkill
- sellersprite-amazon-opsSkill
- top10stats-data-videoSkill
- video-producer-coreSkill
- webbridge-archive-publisherSkill
- web-collection-orchestratorSkill
- wechat-account-data-reviewSkill
- wechat-article-knowledge-packSkill
- wechat-group-insight-reportsSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
18 rules. 0 are common, 15 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
15
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Read project-local AGENTS.md, DESIGN.md, brand rules, screenshots, existing UI, tokens, and component styles before changing UI.
- Distinctive: line 25
Keep the user-facing response brief unless they request the design rationale.
- Very specific: line 33
When the user is evaluating a new DESIGN.md, design Skill, visual prompt, or design-system rule set, compare the same task across the rule …
- Distinctive: line 42
Keep the product content, data, viewport, and test scenario identical across variants.
- Distinctive: line 43
Use the three variants above whenever an old rule set exists.
- Distinctive: line 44
Do not relabel a page-by-page before/after diff as a design-evolution comparison.
- Distinctive: line 58
Do not generate either comparison artifact for a trivial isolated change unless the user explicitly requests it.
- Distinctive: line 62
When a task belongs to an existing product with two or more related UI pages, treat the product as one rendered system before editing an …
- Distinctive: line 66
Keep PageHeader slots consistent: eyebrow, H1, lead, and actions.
- Very specific: line 67
Keep H1 on the product's --fs-h1, --lh-h1, and letter-spacing tokens.
- Distinctive: line 68
Keep the primary action visually identical across related pages.
- Distinctive: line 70
If a page cannot be brought into compliance through shared tokens and components, rebuild it from the product's template instead of adding …
- Distinctive: line 71
Before delivery, render the related pages side by side at the same viewport.
- Distinctive: line 75
Do not turn a local adjustment into a redesign.
- Distinctive: line 76
Do not make every product look like the same warm-neutral, dark-terminal, gradient, glass-card, or three-column template.
- Distinctive: line 77
Do not use color as decoration: accent and chip colors must retain semantic roles.
- Distinctive: line 78
Do not leave empty states, loading states, error states, or keyboard focus as afterthoughts when the UI needs them.
- Distinctive: line 79
Do not copy proprietary assets or layouts from a visual reference.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Frontend Design System
Use references/design-playbook.md as the canonical global design guidance. Read the full reference for a new page, a redesign, a multi-page system, or when the user supplies screenshots or visual references. For a local color, spacing, typography, or component change, read the relevant sections and first inspect the existing project UI, token files, and component styles.
Task Scale
- Local adjustment: preserve the project's visual language and make the narrowest change that fixes the request. Do not redefine the palette or restructure the page.
- Component adjustment: preserve the page, complete the component's spacing, states, accessibility, and responsive behavior.
- New page or redesign: complete internal product understanding, theme extraction, 2-3 visual directions, a chosen direction, token decisions, implementation, and visual QA before delivery.
- Multi-page system: use the four layers in the reference: Tokens, Layout Primitives, Components, and Templates. Fix consistency problems at the highest affected layer.
Workflow
- Read project-local
AGENTS.md,DESIGN.md, brand rules, screenshots, existing UI, tokens, and component styles before changing UI. Project-local rules override this skill. - Determine the product type, audience, density, task scale, and whether the request is local or systemic.
- Derive visual choices from the product's theme and evidence. Do not reuse a familiar palette or layout as a default answer.
- Reuse existing tokens and components when they exist. If a new semantic visual value is necessary, define it at the correct system layer instead of hard-coding it in a page.
- Preserve real content and interaction behavior. Give interactive elements complete states and meet accessibility, responsive, and overflow requirements.
- Render or preview when the environment permits. Check the first viewport, actual content overflow, mobile layout, visible focus, and state changes.
- Keep the user-facing response brief unless they request the design rationale.
Visual Evaluation Artifacts
Choose the comparison artifact that matches the question. Keep every compared output at the same viewport and preserve the screenshots beside the delivered HTML.
Design-evolution comparator
When the user is evaluating a new DESIGN.md, design Skill, visual prompt, or design-system rule set, compare the same task across the rule versions. This is not a before/after page diff.
- Render
Original common output: the task without the new design guidance. - Render
Old DESIGN.md output: the task using the preceding rule set, if one exists. - Render
New DESIGN.md output: the task using the current rule set. - Build one self-contained
design-evolution.htmlthat directly renders and switches between those complete live UI outputs. Its tabs must name the rule variants, not individual page names.
- The comparison surface must render actual DOM and remain interactive. Do not use screenshots, raster images, iframes, or separate HTML documents as the comparison content; use screenshots only for internal visual QA.
- Implement the variants in the same document through a
modestate, shared scenario data, and explicit render functions or framework routes. This preserves offlinefile://behavior and lets the user inspect typography, spacing, controls, focus, and state changes directly. - Keep the product content, data, viewport, and test scenario identical across variants. Change only the design guidance being evaluated.
- Use the three variants above whenever an old rule set exists. If there is no old rule set, compare the baseline and current rules and state that the middle historical variant is unavailable.
- Do not relabel a page-by-page before/after diff as a design-evolution comparison.
Before-and-after comparator
For a redesign, multi-page system migration, or explicit visual-update request that is about implementation changes rather than a rule-set evaluation, capture the original state before editing and the final state after editing. Generate a clickable comparison.html with scripts/create_visual_comparator.py.
python <skill-dir>/scripts/create_visual_comparator.py `
--output <output-dir>/comparison.html `
--title "Updated UI comparison" `
--pair "Workspace|<before.png>|<after.png>"- Repeat
--pairfor multiple pages, or use--manifestwith a JSONpairsarray. - Do not generate either comparison artifact for a trivial isolated change unless the user explicitly requests it.
Cross-Page Integrity Gate
When a task belongs to an existing product with two or more related UI pages, treat the product as one rendered system before editing an individual page.
- Inspect representative sibling pages and identify the shared token file, AppShell, navigation, PageHeader, primary Button, and type scale.
- Require every related page to use the same shared system. A page that uses an alternate top navigation, an alternate sidebar, a page-local palette, a page-local H1 size, or a page-local primary button is outside the system.
- Keep PageHeader slots consistent: eyebrow, H1, lead, and actions. Do not insert page-local KPI blocks, filter callouts, or alternate navigation into the header unless the shared component explicitly supports that variant.
- Keep H1 on the product's
--fs-h1,--lh-h1, and letter-spacing tokens. Do not override these values in page CSS. Scan non-token CSS forfont-sizepixel literals before delivery; token-definition and third-party files are the only permitted exceptions. - Keep the primary action visually identical across related pages. Use ghost or secondary buttons only for subordinate actions, never as a substitute for the page's primary CTA.
- Choose one content-container grammar per section: a true card grid with shared padding and gaps, or a true list with shared row height. Do not mix list separators with card framing.
- If a page cannot be brought into compliance through shared tokens and components, rebuild it from the product's template instead of adding visual overrides.
- Before delivery, render the related pages side by side at the same viewport. Compare AppShell geometry, H1 computed size, PageHeader order, primary-button computed colors, and primary accent token. Treat any unintended difference as a system-layer defect.
Guardrails
- Do not turn a local adjustment into a redesign.
- Do not make every product look like the same warm-neutral, dark-terminal, gradient, glass-card, or three-column template.
- Do not use color as decoration: accent and chip colors must retain semantic roles.
- Do not leave empty states, loading states, error states, or keyboard focus as afterthoughts when the UI needs them.
- Do not copy proprietary assets or layouts from a visual reference. Extract its visual language and recombine it for the current product.
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
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.

