frontend-design
Design, build, restyle, or review polished frontend interfaces for web apps, dashboards, tools, landing pages, components, and responsive workflows.
Skill by AnthonyPoschen
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/anthonyposchen-frontend-design
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design, frontend
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 26, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/frontend-design
What's inside
- Skill name
frontend-design- Description
- Design, build, restyle, or review polished frontend interfaces for web apps, dashboards, tools, landing pages, components, and responsive workflows. Use whenever a request involves frontend visual design, layout, styling, UI hierarchy, interface polish, design systems, color palettes, typography, color schemes, prefers-color-scheme, dark and light themes, responsive behavior, navigation and wayfinding, or making an existing web experience more usable and distinctive. Also use for motion, elevation, accessibility, and design critique. Apply it to both new screens and focused UI improvements, even when the user asks only for code. Avoid templated AI aesthetics. When designing or reviewing UI, visually inspect the running interface with the eyes-on skill, in both dark and light themes.
- License
- MIT
- Instructions
- 241 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- design-brief.md · references/
SKILL.md (first lines)
# Frontend Design
You already know how to design. These are reminders for the places where
agent-built interfaces most often fall short. The brief and the project's
`DESIGN.md` win over anything here. Make each visual choice serve the person
using the page and the page's one job.
## Before You Build
- Name the user, the outcome they need, and the page's single job. Design the
smallest complete workflow first, including its empty, loading, error and
success states, before you choose a shell or navigation.
- When improving existing UI, look at it running first (`eyes-on`). Keep the
conventions that work and change the few things with the most impact.
- Ask only when an unknown would change the workflow, content priority, or
visual direction. Otherwise state the assumption and proceed.
- Write a short plan before a new screen or redesign: the direction in a few
concrete words, the structural system, the type voices and their jobs, the
colour roles, and the one signature element you will spend boldness on.
- Choose the direction from the subject and the brief, not from habit. When
the brief pins a direction, execute it fully and precisely. When an axis is
free, ask whether your choice would appear unchanged for an unrelated
product; if it would, find a reason in this subject or change it.
## Keep A Design Record
`DESIGN.md` in the target project is the durable record of design decisions.
Read it before a material decision. Create it at the first substantial design
consultation, and update it when feedback, testing or review changes a
decision. Record the decision and its reason, not the conversation. Use
[the `DESIGN.md` template](references/design-brief.md) for a substantial flow
or redesign. Project palettes, typefaces and signature elements belong there,
not in this skill.
When someone is choosing a palette, let them compare a few named candidates on
a real screen of the product (a local-only switcher), then keep only the
winner's tokens.
## Structure
- Give the content a structural system to sit in: a column, a grid, rules or
a spacing rhythm, and make every section obey it, with the same spacing
between sections and no stray empty bands. Visible structure can carry a
page with very little decoration. Hold one shape language (corner radius,
border weight) across every component, chips, tags and badges included.
- When items repeat (cards, rows, plans), every item shows the same fields in
the same order, so people compare instead of re-reading.
- The primary action matches the page's job and leads to the thing the page
is about. On a long page, repeat it at the end. Keep the first view lean:
the page's job, the primary action and the key overview stay above the
fold; anything else (contact, sharing, social, extra facts) goes lower.
- Give persistent information (navigation, current location, status) a
stable position, and keep sticky or fixed chrome from covering content,
headings or focused elements. Offset in-page targets for it.
- Keep text at a readable measure. On wide screens, use the spare width for
navigation or deliberate space, not longer lines.
- A loading, empty or refreshed state keeps the settled state's box. Nothing
the user can select moves when data arrives. Media reserves its aspect
ratio.
- Feedback and indicators never reflow the text they describe.
## Type
- Use a small number of type voices, each with one job (for example display,
reading, and a utility voice for labels, data and controls). Give each text
role one repeatable treatment. When the brief or design record assigns a
voice to a role (navigation, tags, buttons), apply it, casing included,
everywhere that role appears: side navigation, tags, secondary controls and
narrow layouts.
- Small labels that carry meaning still pass contrast. Make them quieter with
size, weight or case, not faint grey. Keep the faintest tones for rules,
ticks and decoration.
- When a label does not fit, shorten the words. Do not truncate meaning.
- Headings add information beyond a section's number or label rather than
repeating it. Keep multi-line headings from stranding one short word.
## Colour
- Keep most of the surface neutral and give colour a job: action, state, or
identity. One family per job.
- State colours mean state and stay the same on every item. Identity colours
identify. Never let an identity colour read as a warning or error, and never
colour state by identity.
- When sibling items each get an identity colour, choose the set together and
check it side by side in both themes. Every pair should read as clearly
different at a glance (aim for roughly ΔE 25), apart from the state colours,
and saturated enough to read as a colour, not as ink. Neighbours in a
spectrum (blue, indigo, violet) read as one family. When hue separation runs
out, separate by lightness, shape, label or placement rather than forcing
more hues. An item's identity (colour, type voice) carries everywhere the
item appears: overview strips, navigation and its own card. Where it colours
text, darken the light-theme variant until that text passes.
- Colour is never the only signal for status, selection or errors. When
colour, shape or size encodes something that isn't self-evident, say what it
means on the page with a short key or legend.
- Build both a light and a dark theme and follow `prefers-color-scheme`. If
there is a theme toggle, remember the choice and switch without a flash.
- Design the dark theme rather than inverting it: lift accents so they hold
contrast on dark, flip the primary emphasis, keep hairlines dim but
visible, raise surfaces instead of relying on shadows, then recheck every
label against the accessibility bar.
## Wayfinding
- On a long page or flow, help people answer: how far am I, where am I, and
what comes next. One element can carry progress, contents and navigation
together.
- Navigation aids carry real information and are operable. Section names in
them are real links whose labels match the headings, the current one is
marked (`aria-current`), and they sit early in the tab order. Any control
that moves the reader (a progress scale, a minimap) also has a keyboard
path. Do not draw an indicator that looks interactive and is not.
- When you build a section rail or contents list, make sure it includes:
previous and next placed around the list, as links like the section items;
the state of each section shown with words ("here", "next", "read") or
proportion, not only a paler tone; every label at 4.5:1 or better; and after
a jump, focus moved into the target section so keyboard and screen-reader
users land where they asked to go.
- Keep numbering and names consistent between headings, navigation andFrom AnthonyPoschen/agent-skills, MIT.
More from AnthonyPoschen/agent-skills
- 0.603score
scripted-verification
AnthonyPoschen
Use when you are about to read logs, screenshots, diffs, timings, or benchmark numbers to decide whether something passed; when writing a "check", "verify"…
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
coding
AnthonyPoschen
ALWAYS activate for programming-language work and code-adjacent project work: implementation, refactoring, debugging, code review, tests, CI fixes…
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
create-verification-skill
AnthonyPoschen
Create a project-local verification skill by reading the codebase.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
eyes-on
AnthonyPoschen
Look at a rendered change the way a person would, before calling it done.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
dockerfile
AnthonyPoschen
Create, review, or update Dockerfiles and container build workflows.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
git
AnthonyPoschen
Git commits, worktrees, and releases.
SkillAFreshness: A★ 2Claude Code +9 - 0.598score
in-your-own-words
AnthonyPoschen
Use before starting ambiguous or multi-step work, when the user says "in your own words" or "restate my goal", or when intent is unclear.
SkillAFreshness: A★ 2Claude Code +9 - 0.598score
lint-rules
AnthonyPoschen
Use when fixing a bug, the user corrects the same mistake twice or says "you missed some", a review comment or gotcha recurs, or you are about to write a…
SkillAFreshness: A★ 2Claude Code +9 - 0.598score
maintain-verification-skill
AnthonyPoschen
Keep a project's verification skill and feature map honest.
SkillAFreshness: A★ 2Claude Code +9 - 0.598score
makefile
AnthonyPoschen
Use when creating, reviewing, or editing Makefiles, make targets, make recipes, task runners, or documentation that tells users to run make commands.
SkillAFreshness: A★ 2Claude Code +9 - 0.598score
omarchy-development
AnthonyPoschen
Develop, test, debug, or contribute to Omarchy source code in an isolated environment.
SkillAFreshness: A★ 2Claude Code +9 - 0.598score
flux
AnthonyPoschen
FluxCD GitOps work.
SkillAFreshness: A★ 2Claude Code +9
Signals
- Stars
- 2
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 10, 2026 (19 days ago)
- Last release
- Oct 14, 2026 (14 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/AnthonyPoschen/agent-skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from AnthonyPoschen
- 0.603score
coding
AnthonyPoschen
ALWAYS activate for programming-language work and code-adjacent project work: implementation, refactoring, debugging, code review, tests, CI fixes…
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
git
AnthonyPoschen
Git commits, worktrees, and releases.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
create-verification-skill
AnthonyPoschen
Create a project-local verification skill by reading the codebase.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
dockerfile
AnthonyPoschen
Create, review, or update Dockerfiles and container build workflows.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
eyes-on
AnthonyPoschen
Look at a rendered change the way a person would, before calling it done.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
scripted-verification
AnthonyPoschen
Use when you are about to read logs, screenshots, diffs, timings, or benchmark numbers to decide whether something passed; when writing a "check", "verify"…
SkillAFreshness: A★ 2Claude Code +9
Cite this page
frontend-design skill. Agent Mart. https://agentmart.dev/skills/anthonyposchen-frontend-design (last updated 2026-10-24)