agentleFS
igovetigovet/cortex-orch-acp/skills/design-frontend-design/SKILL.md

design-frontend-design

Skill1 starsChanged 59 days agoMITView on GitHub

Distinctive, intentional visual design for new or reshaped user-facing UI, covering type, palette, layout and restraint so the result does not read as a templated default. Use at the implementation stage when visual choices must be made or an approved design direction must be built.

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 14 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈14

    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:

    Use this skill when a user-facing screen or component must be built or reshaped and its visual choices matter.

    line 8
  2. Distinctive:

    If the brief pins a visual direction, a palette, a font or an era, follow it exactly, including when it asks for a look this skill would …

    line 26
  3. Distinctive:

    Plan: Write a compact token plan:

    line 33
  4. Distinctive:

    Review the plan against the brief: If any part reads like the default you would produce for any similar page, revise it and say what …

    line 39
  5. Distinctive:

    Only then write code.

    line 41
  6. Distinctive:

    When writing CSS, check selector specificity.

    line 43
  7. Distinctive:

    Use one family, or two that are clearly distinct.

    line 51
  8. Distinctive:

    Keep line length under about 80 characters.

    line 53
  9. Distinctive:

    Treat a headline type treatment as an active design element, not a neutral container for the text.

    line 55
  10. Distinctive:

    Use one orchestrated motion moment (for example one page-load sequence or one reveal) rather than scattered fades on every section and …

    line 82
  11. Distinctive:

    Write copy as design content.

    line 85
  12. Distinctive:

    Verify the rendered result, not only the source:

    line 99
  13. Distinctive:

    If no capture tool is available, say so in the report.

    line 105
  14. Distinctive:

    Do not rate your own design (for example "9/10") and treat the rating as proof of quality.

    line 113

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Frontend Design

Use this skill when a user-facing screen or component must be built or reshaped and its visual choices matter. Ordinary frontend work with an approved design direction still follows that direction; this skill then adds only the craft rules below.

Adapted for Cortex from Anthropic's frontend-design skill (Apache-2.0; see LICENSE.txt and CORTEX-SOURCE.md).

Ground the design in the subject

Identify the product, its audience and the screen's primary job before choosing any visual treatment. If material facts are missing, use your advertised ask_question route rather than inventing the subject, audience or primary job. Label only non-material implementation assumptions. Distinctive choices come from the subject's own materials, vernacular and context. A children's reading app and a financial analysts' dashboard should not share a visual language.

If the brief pins a visual direction, a palette, a font or an era, follow it exactly, including when it asks for a look this skill would otherwise avoid.

Plan, review, then build

Work in two passes.

  1. Plan. Write a compact token plan:
    • Color: 4 to 6 named values with their roles.
    • Type: one or two families, their roles, and a clear type scale.
    • Layout: one concept, described in a sentence, with alignment guidance (left, center or justified) and a simple ASCII sketch if it helps compare options.
    • Principles: the two or three rules that make this screen specific.
  2. Review the plan against the brief. If any part reads like the default you would produce for any similar page, revise it and say what changed and why. Only then write code.

When writing CSS, check selector specificity. Type selectors and class selectors can cancel each other's padding and margin, especially between sections.

Typography

  • Typography carries the page's character. Choose typefaces for this brief, not the families you reach for on every project.
  • Use one family, or two that are clearly distinct. A display face and a body face that look alike add noise.
  • Keep line length under about 80 characters. Serif body text can run slightly longer and needs slightly more line height than sans-serif.
  • Treat a headline type treatment as an active design element, not a neutral container for the text.

Avoid the common tells

Apply the hard bans in the banned AI-tell patterns (read_skill id design-banned-patterns): no one-sided decorative accent stripes, and no decorative leading dots or default meta-chips. Only a product-spec-required semantic status indicator is permitted; a dot is never its only signal.

These treatments mark generated pages. Avoid them unless the brief asks for them:

  • Accenting one word of a headline with italic, bold or a second color.
  • All-caps labels, especially tracked-out eyebrow labels above every heading.
  • Typographic labels added above content with no informational role.
  • Middle-dot meta strings ("A · B · C") and "WORD — fragment" labels.
  • Numbered markers (01, 02, 03) unless the content is truly a sequence.
  • A single warm cream background with a terracotta accent, or a near-black background with one acid accent, used by default.
  • Identical rounded cards with one radius and one soft shadow for everything, and gradient washes used as decoration.
  • A "big number, small label, gradient accent" stat block used by default.
  • A monospace face for small data labels, or a trailing arrow on every link.

Structural devices such as rules, borders, dividers and labels should encode information. If removing one changes nothing the reader needs, remove it.

Motion, copy and restraint

  • Use one orchestrated motion moment (for example one page-load sequence or one reveal) rather than scattered fades on every section and hover effects on every card. Motion that answers a user action and shows what changed is welcome.
  • Write copy as design content. Name things in the user's words, use active voice for actions ("Save changes", not "Submit"), keep errors specific about what happened and how to fix it, and make empty states an invitation to act.
  • Spend boldness in one place. Let one element be the memorable thing and keep the rest disciplined. Cut any decoration that does not serve the brief.

Quality floor

Build to this floor without announcing it: responsive from phone width to desktop, visible keyboard focus, reduced motion respected, text and control contrast that meets WCAG AA, and a harmonious palette with semantic color tokens.

Rendered verification

Verify the rendered result, not only the source:

  • Capture desktop and mobile widths when a capture tool is available in this session (a browser tool if one is listed, or the project's own test tooling).
  • Open each captured image and note concrete defects, with the region and the width where each appears.
  • If no capture tool is available, say so in the report. Rendered claims then stay unverified, and the work goes to review with that limitation stated.

Hand the screenshot paths and the list of open visual risks to the review stage (see design-visual-qa).

No self-scoring as evidence

Do not rate your own design (for example "9/10") and treat the rating as proof of quality. Quality claims rest on opened screenshots, concrete findings, and re-checks after changes. A number without those is not evidence.

Output

Report the token plan, the revisions made after the review pass, the files changed, the screenshots opened (paths and widths), the open visual risks, and what was not verified.

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.