agentleFS
rahmanef63rahmanef63/si-coder-agent/skills/sc-ui/SKILL.md

sc-ui

Skill13 starsChanged 2 months agoMITView on GitHub

SI-Coder visual interface quality skill. Design, audit, and refine web UI with strong hierarchy, typography, spacing, color, density, motion, component states, responsive craft, and an anti-AI-slop filter while preserving an existing product's design DNA by default.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep durable instructions in English.

    line 15
  2. Very specific:

    Before editing, produce an internal Design Read from the available product/code/rendered UI:

    line 20
  3. Distinctive:

    Prefer a small coherent system over many local decisions:

    line 54
  4. Distinctive:

    Never redesign or modify an explicitly excluded surface.

    line 67
  5. Distinctive:

    Do not call visual work done from code inspection alone.

    line 70
  6. Very specific:

    no accidental overflow/clipping/wrapping,

    line 72
  7. Distinctive:

    no new generic/slop patterns introduced.

    line 78

≈ “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

sc-ui — visual craft without AI slop

Language

Keep durable instructions in English. Reply in the user's language unless requested otherwise.

Default behavior

Inspect before inventing. If the project already has a coherent UI, preserve its design DNA unless the user explicitly asks for a redesign. Do not reset an existing product to generic defaults.

Before editing, produce an internal Design Read from the available product/code/rendered UI:

  • product type and primary jobs,
  • audience and expertise,
  • platform and viewport priorities,
  • hierarchy and information density,
  • typography and spacing rhythm,
  • surface/border/radius/shadow language,
  • color roles and contrast,
  • icon and illustration language,
  • motion language,
  • existing design tokens/components,
  • explicit exclusions from the user.

Anti-slop filter

AI slop is not a specific color or component. It is unearned sameness: choices made because they are common in generated interfaces rather than because they serve the product.

Apply a purpose test to every visible decision: What user/product job does this choice serve?

Flag and fix when unjustified:

  • cards wrapping every section or nested cards with no hierarchy benefit,
  • decorative gradients/glass/blur/noise that do not communicate state or brand,
  • excessive pills, badges, rounded rectangles, icon buttons, and floating containers,
  • generic centered hero/marketing composition inside an application shell,
  • fake metrics, fake social proof, placeholder dashboards, decorative charts, or meaningless activity,
  • repeated same-sized modules that flatten hierarchy,
  • arbitrary huge headings, low-information whitespace, or over-spacious admin/tool UIs,
  • excessive shadows/radii that erase grouping structure,
  • animation on everything rather than on state change or spatial continuity,
  • hard-coded one-off colors/spacing where project tokens already exist,
  • rebuilding a mature design system instead of extending it.

These are detectors, not blanket bans. Keep any pattern that has a clear product, accessibility, or brand reason.

Visual system contract

Prefer a small coherent system over many local decisions:

  1. Typography: clear roles for display/title/body/label/code; readable measure and line height.
  2. Spacing: repeatable rhythm; proximity communicates grouping before boxes do.
  3. Color: semantic roles before decorative palette; preserve brand tokens.
  4. Shape: radii/borders/shadows have explicit hierarchy roles.
  5. Density: match task frequency and expertise; tools can be compact, reading surfaces can breathe.
  6. Components: all interactive states exist — default, hover where relevant, focus-visible, active/selected, disabled, loading, error.
  7. Motion: communicate causality, continuity, feedback, or hierarchy; respect reduced-motion preferences.
  8. Responsive behavior: preserve task priority rather than merely stacking desktop rectangles.

Desktop shell rules

For application/workbench shells, prefer frame → navigation → primary work area → contextual tools/inspector over dashboard-card composition. Keep frequent actions visible, secondary actions discoverable, separators aligned, resize behavior intentional, and selected/focus states unmistakable.

Never redesign or modify an explicitly excluded surface. If the user says not to touch mobile navigation, treat it as a hard scope lock.

Verification gate

Do not call visual work done from code inspection alone. When tools allow it, inspect rendered pixels at representative sizes and verify:

  • hierarchy is obvious in a 3-second scan,
  • no accidental overflow/clipping/wrapping,
  • alignment and divider continuity,
  • typography and density consistency,
  • component states,
  • light/dark theme if supported,
  • responsive behavior,
  • no new generic/slop patterns introduced.

Report concrete defects and fixes, not subjective praise.

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.