swarm-design-ui
Skill5 starsChanged 4 months agoMITView on GitHub
UI/UX design phase — propose an intentional design system (pattern, brand, color, typography, spacing, atomic components) with the user, offering distinct directions and per-page layout options, for any interface type: Web, Mobile, Desktop, TUI, or CLI. Use when the SRS declares a user interface, when defining a design system or visual identity, choosing layouts, or reviewing UI work against UX laws.
In this repository14 other files it gives its agents.
- swarm-debugSkill
- swarm-design-deepSkill
- swarm-designSkill
- swarm-ejectSkill
- swarm-flowSkill
- swarm-implementSkill
- swarm-initSkill
- swarm-migrateSkill
- swarm-orchestrateSkill
- swarm-reviewSkill
- swarm-skill-forgeSkill
- swarm-specSkill
- swarm-vaultSkill
- swarm-writeSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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
- Very specific: line 25
Don't jump to one look. Using the product's category and personality (see references/style-menu.md for the vocabulary), synthesize a …
≈ “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
swarm-design-ui — UI/UX design phase
Gate: validated SRS declaring a user interface. Runs alongside swarm-design. The goal is intentional, consistent design decided with the user — not templated defaults.
Load only what you need (token economy): the per-interface reference for the type at hand
(references/web.md, references/mobile.md, references/desktop.md, references/tui-cli.md),
and — for graphical UIs — references/style-menu.md, the on-demand catalog of product
categories, style families, layout patterns, and design resources (checklist.design, UX laws,
color/type/psychology tools).
1. Interview first
Audience and context of use; brand personality; platform conventions to honor; accessibility requirements; existing brand assets. Keep it light and options-first — the point is to understand the product, not to quiz.
2. Propose a design system (offer ≥2 directions)
Don't jump to one look. Using the product's category and personality (see
references/style-menu.md for the vocabulary), synthesize a recommended design system and
at least one distinct alternative — e.g. "clinical precision" vs "warm craft" — so the user
chooses with something concrete in front of them. Present each as a compact panel:
DESIGN SYSTEM — <recommended | alternative>
Pattern/layout : <e.g. Hero-centric + social proof> why: <conversion/trust rationale>
Style : <e.g. Soft UI evolution> best for: <fit>
Colors : primary / surface / accent / semantic / bg / text (WCAG-AA checked)
Typography : <display / body pairing> mood: <…>
Effects : <shadows, motion timing, hover>
Avoid : <anti-patterns for this domain — see style-menu>The style/category menu is a guide, not a cage: if the product's needs point elsewhere, design for them. Resolve the picked direction into the tokens below.
3. Produce the design-system doc (docs/design-ui.md)
Everything as tokens and scales, never per-screen ad-hoc values:
- Color — palette with roles (background/surface/primary/accent/semantic), light + dark, WCAG-AA contrast checked.
- Typography — 1–2 families, a modular size scale, line-height and weight rules.
- Spacing & box model — one spacing scale (e.g. 4px base), consistent margin/border/padding discipline, grid + auto-layout container rules.
- Atomic inventory — atoms → components → templates → pages; every screen implied by an FR appears here, traced to its FR-ID. Use the correct component names (namethatui.com is the dictionary): "drawer", "segmented control", "combobox" — not "that sliding panel". Precise names in specs and tickets mean every agent builds the same thing; resolve vague descriptions to the proper name and confirm.
- Per-page layout — for each key screen, offer a couple of layout options
(
references/style-menu.mdlists landing-page and dashboard patterns) and record the choice with its rationale, not just a single default. - Responsive rules — breakpoints, and a genuinely distinct mobile design where the SRS calls for mobile (not a shrunken desktop).
- Motion & interaction — states (hover/focus/active/disabled/loading/error), timing.
- Voice — the design system is half the identity; how the product sounds is the other
half. Every string in the inventory (labels, errors, empty states, tooltips, onboarding)
goes through swarm-write, against the project's
voice.md. Copy is interface, not decoration applied afterwards.
4. Review against the UX-laws checklist (apply as review, not jargon)
Fitts (targets big and near) · Hick (fewer choices at decision points) · Jakob (follow
platform conventions) · Miller (chunk information) · proximity/common-region (group
related things) · Doherty (feedback < 400 ms) · error prevention over error messages. Run the
pre-delivery checklist in references/style-menu.md before calling UI work done.
Rules
- Accessibility is explicit for graphical UIs: contrast ratios, focus order, touch targets, reduced-motion.
- TUI/CLI projects: skip all graphical guidance and the style menu; help text, flags, exit
codes, and error style ARE the design (see
references/tui-cli.md). - End with user validation (gated) or FR-coverage self-check (auto); update flow-state.
Influences: Laws of UX (Yablonski); Atomic Design (Frost); roadmap.sh/design-system; checklist.design; Figma typography guide; NameThatUI (component vocabulary); Anthropic frontend-design skill — see CREDITS.md.
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.

