agentsclimarketplace

Ux

Skill justinramos101/agent-skill-kit/skills/ux

Battle-tested Agent Skills for coding agents — source-grounded, failure-driven heuristics that audit and design real surfaces. Install with npx skills.

Install
npx -y skills add justinramos101/agent-skill-kit --skill ux

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

End-user UX and UI, both directions: audit what exists, design what doesn't. Audit: 'UX review', 'accessibility audit', 'is this WCAG compliant', heuristic evaluation of a confusing flow, form friction and drop-off, navigation/IA review, error-and-recovery copy, keyboard/screen-reader/contrast passes. Design: 'design this dashboard/UI', build a screen, prototype, or slide deck, author a design system with tokens, add motion, elevate a working-but-generic app to feel premium ('add polish', 'make it feel premium', 'elevate the whole app'), run an 'anti-slop UI pass' on generated UI, prep an artifact for host embedding or export. Do NOT use for a go-all-out maximum-creativity showpiece where distinctiveness is the whole brief, retention or re-engagement strategy on its own, developer-facing surfaces (APIs, SDKs, CLIs, error envelopes), docs as a reading surface (READMEs, quickstarts), UX of surfaces an agent operates, or prose editing.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.9 KB, as published. Nobody here has run it

UX

Infer direction from the request — never ask which mode:

  • Audit an existing surface → usability flow, accessibility/WCAG, form, navigation/IA, or error-recovery pass. Load references/auditing.md.
  • Design or build UI → screen, dashboard, prototype, design system, deck, or motion; the anti-slop polish pass; or elevate a whole working app to feel premium (craft, rationed signature moments, retention loops under an ethics gate). Load references/crafting.md.
  • Artifact lives inside an embedding host (tweak panel, direct edit, speaker notes, export)? Also load references/host-integration.md.
  • Before calling a built artifact done, run references/browser-verification.md — reading source is not seeing the render.

Core principle: the next correct action must be visible, reversible, and accessible. Audits find where it isn't; new UI is designed so it is — context before craft: read the product, existing system, and assets before inventing visuals.

Invariants

  • WCAG is the floor, not the ceiling — passing a checker does not make a flow usable. Automated scanners catch only a minority of failures: keyboard path, focus order, semantics, contrast at zoom, and assistive-tech behavior need human checks, so every accessibility finding separates likely WCAG failures from issues needing manual verification and carries a manual-verify note.
  • Anti-slop bar. Generated UI that could have come from a generic AI template is not done. One stock trope may be intentional; three means restart. Name the committed direction — material, type voice, accent, density — in the handoff.
  • Honest content. No fabricated metrics, logos, or testimonials — explicit placeholders instead. Never silently invent a design-system component. Respect prefers-reduced-motion. No dark-pattern retention mechanics (streak-shaming, fake urgency); don't clone an identifiable brand's UI or visual identity — extract principles into an original design.

Workflow

  1. Assess. Audit or build? Which slice — flow, accessibility, form, navigation, errors / screen, system, prototype, deck, motion, host? State the inference; proceed.
  2. Load only the matching reference(s).
  3. Apply the named heuristics. Each audit finding names the affected user, the heuristic, and a concrete fix; each design choice states its commitment and trade-off.
  4. Emit. Audit: findings ordered by user impact, with verification per finding. Build: a runnable artifact plus stated assumptions and variation axes worth exploring — verified in a browser before handoff.

Run to completion; state assumptions inline. Ask only when the request is truly directionless and the choice consequential. Stop after one emission — a second pass only if browser verification or the accessibility checks flag a miss; never a third.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.