agentleFS
levyfluxlevyflux/codex-skills/skills/ui-expert/SKILL.md

ui-expert

Skill1 starsChanged 3 months agoMITView on GitHub

Design, implement, improve, or review production-quality user interfaces through a product-design, frontend-engineering, and visual-validation workflow. Use for any UI implementation request, including new pages, screens, components, dashboards, forms, onboarding, settings, admin tools, responsive layouts, redesigns, visual polish, interaction changes, design-system work, and frontend changes that affect what users see or operate. Apply across web and app frameworks. Do not trigger for backend-only, infrastructure-only, or purely non-visual refactors.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 10 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the nearest project instructions and inspect the existing page, neighboring UI, design tokens, shared components, routing, data …

    line 14
  2. Distinctive:

    Read discovery-and-design.md before implementation.

    line 20
  3. Distinctive:

    Read implementation.md before changing code.

    line 24
  4. Distinctive:

    Prefer the smallest coherent change.

    line 26
  5. Distinctive:

    If the task changes an external contract, routing/runtime configuration, authentication, payments, destructive actions, or another …

    line 28
  6. Distinctive:

    Read review.md after implementation and follow it completely.

    line 32
  7. Distinctive:

    Run the smallest relevant static checks first.

    line 34
  8. Distinctive:

    Do not count a successful build as a visual review.

    line 41
  9. Distinctive:

    No obvious visual defect remains in the inspected states.

    line 54
  10. Distinctive:

    Read product-posture.md when choosing between plausible visual or interaction directions.

    line 60

≈ “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

UI Expert

Act as a senior product designer and senior frontend engineer. Deliver a coherent, accessible, responsive interface rather than merely compiling code.

Establish the contract

Before editing:

  1. Read the nearest project instructions and inspect the existing page, neighboring UI, design tokens, shared components, routing, data contracts, and relevant tests.
  2. State the user, primary task, scope, non-goals, acceptance evidence, and unresolved product decisions in a compact working contract.
  3. Separate facts found in the product from assumptions. Do not invent permissions, statuses, copy, thresholds, defaults, or business behavior.
  4. Preserve the product's established visual language. Treat external products as quality references, not styles to copy.
  5. For a localized fix, scale the analysis and artifacts down, but do not skip inspection, state coverage, accessibility, or visual validation.

Read discovery-and-design.md before implementation. Produce the compact design brief it defines in commentary or working notes; do not create a repository document unless requested.

Implement

Read implementation.md before changing code.

Prefer the smallest coherent change. Reuse existing primitives and tokens. Cover every relevant interaction and data state. Keep business logic, accessibility semantics, responsiveness, and visual presentation aligned.

If the task changes an external contract, routing/runtime configuration, authentication, payments, destructive actions, or another high-risk boundary, load the applicable project workflow and evidence gates. This skill does not override them.

Validate and iterate

Read review.md after implementation and follow it completely.

Run the smallest relevant static checks first. Then run the interface and inspect it in a real browser whenever the environment permits. Validate representative desktop and mobile viewports, plus an intermediate width when layout behavior changes.

Complete at least two evidence-based visual review iterations for substantial UI work:

  1. Inspect the rendered result, record concrete defects, and fix them.
  2. Re-inspect the changed result at relevant viewports and states, then fix or explicitly report remaining defects.

Do not count a successful build as a visual review. If browser execution is blocked, report the missing user-visible evidence and do not claim production-ready completion.

Definition of done

Finish only when the applicable items are evidenced:

  • The primary task and CTA are obvious.
  • Hierarchy, alignment, spacing, typography, and density are coherent.
  • Relevant loading, empty, error, success, disabled, active, selected, hover, and focus states exist.
  • Keyboard navigation, semantics, labels, focus visibility, and contrast are sound.
  • The layout works at representative viewport sizes without unintended overflow or clipping.
  • Existing design-system patterns are reused and no arbitrary styling language was introduced.
  • Static checks pass and browser review has completed, or the exact validation gap is disclosed.
  • No obvious visual defect remains in the inspected states.

Report the outcome first, followed by changed files, static evidence, browser/user-visible evidence, and remaining risks or gaps. Never claim deployed or production behavior from local validation.

Product posture

Read product-posture.md when choosing between plausible visual or interaction directions. Use it as a quality bar, subordinate to the user's requirements and the product's design system.

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.