design-showcase
Skill1 starsChanged 8 months agoMITView on GitHub
Build a side-by-side visual comparison of 3-7 design variants for a component or view. Dispatches a single Opus agent that produces a self-contained HTML showcase, serves it locally, and lets the user pick a direction. Use when the user wants to explore design options before committing to an implementation.
In this repository51 other files it gives its agents.
- algorithmic-artSkill
- audit-componentsSkill
- canvas-designSkill
- cascade-orchestrationSkill
- consultSkill
- data-reconciliationSkill
- debug-collaborateSkill
- deep-root-causeSkill
- deploy-allSkill
- design-reverse-engineerSkill
- docxSkill
- env-auditSkill
- evolveSkill
- frontend-designSkill
- full-stack-buildSkill
- full-stack-traceSkill
- gh-repo-reconSkill
- grill-meSkill
- handoffSkill
- html-slidesSkill
- insightSkill
- internal-commsSkill
- isolation-testSkill
- mcp-builderSkill
- model-selectionSkill
- pdfSkill
- persistent-issueSkill
- pre-merge-reviewSkill
- premise-auditSkill
- propagate-featureSkill
- regression-bisectSkill
- research-gateSkill
- review-implSkill
- schema-checkSkill
- session-notesSkill
- skill-auditSkill
- skill-creatorSkill
- slack-gif-creatorSkill
- start-daySkill
- temporal-forensicsSkill
- theme-factorySkill
- throwaway-scriptSkill
- triage-ideasSkill
- verify-completeSkill
- webapp-testingSkill
- web-artifacts-builderSkill
- wisdomSkill
- worker-buildSkill
- worktree-guardSkill
- write-planSkill
- xlsxSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 7 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 28
Read the component file and its CSS.
- Distinctive: line 38
Write one consistent fake dataset that all variants use.
- Very specific: line 72
Use model: "opus" — this is creative/design work and the quality difference is significant.
- Distinctive: line 93
No React, Vue, or Angular — vanilla HTML + CSS + minimal JS
- Distinctive: line 95
Must include a "Current" control baseline so users have an anchor
- Very specific: line 100
Use <template> for index.html structure — see template.html in this directory
- Distinctive: line 104
After the user picks a variant:
- Distinctive: line 106
Do NOT continue in the showcase agent — open a new task
- Distinctive: line 107
Read the picked variant's HTML section to extract the design decisions
- Very specific: line 109
Follow /frontend-design for token usage and component patterns
- Very specific: line 110
Run npm run build to verify before deploying
≈ “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.
Files
Design Showcase
Produces a self-contained HTML page with multiple design variants rendered side-by-side. User reviews, picks a direction, then a follow-up implementation pass converts the chosen variant into real code.
When to use
- Component or view redesign with unclear direction
- Exploring a new design philosophy or pulling inspiration from a reference site
- "Show me options before we commit" requests
- Design reviews where stakeholders need a visual to react to
When NOT to use
- Minor CSS tweaks (color swap, spacing adjustment) — just make the change
- Single-option tasks where the direction is already decided
- Backend-only work
- Feature flags / A-B testing in production code
The Workflow
Step 1 — Read the current component
Read the component file and its CSS. Capture the exact data shape it renders. Note: what information is shown, how it is grouped, what actions exist.
Step 2 — Define 3-7 design philosophies
Pick from philosophies.md or invent new ones for the task. Each variant needs a distinct visual identity — if two variants look similar, merge them or push them further apart. Always include a "Current" baseline (Variant 00) so the user has a reference anchor.
See philosophies.md (in this directory) for the full catalog.
Step 3 — Invent realistic fake data
Write one consistent fake dataset that all variants use. For data-heavy components: use real-sounding identifiers, descriptions, quantities, and dates. Consistent data is critical — variants must differ only in design, not in content.
Step 4 — Dispatch one Opus agent
Agent({
model: "opus",
isolation: "none", // read-only, no code changes
prompt: `
Build a self-contained HTML design showcase at c:/tmp/<slug>-showcase/index.html.
Component: <name>
Fake data: <dataset>
Variants to include:
00. Current — faithful recreation of the existing design
01. <Philosophy 1 name> — <one-line description>
02. <Philosophy 2 name> — <one-line description>
...
[Paste philosophy specs from philosophies.md]
Requirements:
- Single HTML file, no build step, no external JS frameworks
- Google Fonts via CDN is OK
- Top nav with anchor links to each variant (#variant-00 through #variant-N)
- Dark-mode toggle in the nav bar (CSS class toggle, no JS frameworks)
- Each variant in its own full-width section with a header card showing the variant name and 2-3 bullet design rationale
- Responsive down to iPad (768px)
- Include a "Conjunction" blend variant as the last entry
`
})Use model: "opus" — this is creative/design work and the quality difference is significant.
Step 5 — Serve and present
cd c:/tmp/<slug>-showcase && npx serve . -p 4321Direct the user to http://localhost:4321. Ask them to review and call out which variant number they prefer, or which elements they want to combine.
Output convention
c:/tmp/<component-name>-showcase/
index.html # the showcase (self-contained)Slug convention: info-modal, kpi-card, data-table, entry-form.
HTML scaffold guardrails
- No React, Vue, or Angular — vanilla HTML + CSS + minimal JS
- Google Fonts via CDN is allowed (
<link rel="preconnect">+ font link) - Must include a "Current" control baseline so users have an anchor
- Variants must be visually distinct — if two look the same, cut one or push them apart
- Dark-mode toggle at top (toggle
data-theme="dark"on<html>) - Responsive to iPad (768px minimum)
- Each section includes a rationale card (what principles guided this variant)
- Use
<template>forindex.htmlstructure — seetemplate.htmlin this directory
Handoff after pick
After the user picks a variant:
- Do NOT continue in the showcase agent — open a new task
- Read the picked variant's HTML section to extract the design decisions
- Implement those decisions in the real component (React + your project's CSS conventions)
- Follow
/frontend-designfor token usage and component patterns - Run
npm run buildto verify before deploying
The showcase is a design artifact, not production code. Treat it as a spec, not a source to copy.
Example prompt template
Redesign the <ComponentName> component. The current design is the baseline.
Produce a showcase at c:/tmp/<slug>-showcase/index.html with these variants:
00. Current — faithful recreation (baseline reference)
01. Impeccable — anti-pattern discipline, sharp corners, IBM Plex Sans
02. UI/UX Pro Max — operator-first, traffic-light urgency, 44pt touch targets
03. Taste — serif heading, monospace numbers, one accent color
04. Linear — zero-chrome, keyboard shortcuts, tight 4px grid
05. 21st.dev — same layout, CSS-only micro-interactions
06. Conjunction — best-of-all with rationale card
Fake data: <paste dataset here>
Philosophy specs: <paste from philosophies.md>
Requirements: single HTML file, Google Fonts CDN OK, dark-mode toggle, anchor nav, iPad-responsive, rationale card per variant.Related skills
/frontend-design— design tokens and component patterns (use after picking a variant)/webapp-testing— Playwright visual verification after implementing the chosen variant/review-impl— adversarial review before merging the implemented variant
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.

