a11y-audit
Audit a UI or design against WCAG 2.2 AA/AAA and ARIA patterns, returning…
Scaffold a new design-product project that matches the recommended Claude Code layout (the reference structure). Use when starting a fresh product/app that will be built with this design system.
$ npx -y skills add plugin87/ux-ui-agent-skills --skill scaffold-project --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/scaffold-projectContext preview
The summary Claude sees to decide when to auto-load this skill.
Scaffold a new design-product project that matches the recommended Claude Code layout (the reference structure). Use when starting a fresh product/app that will be built with this design system.
name: scaffold-project description: Scaffold a new design-product project that matches the recommended Claude Code layout (the reference structure). Use when starting a fresh product/app that will be built with this design system. # A command the user runs. Converted from .claude/commands/ so it has a # CLAUDE_SKILL_DIR of its own and its paths resolve on every install route. # disable-model-invocation keeps the behaviour a command had: the user # starts it, the model never does, and its description stays out of the # listing budget the model's skill choice is drawn from. disable-model-invocation: true
> **Step 0 — is the kit here?** This skill reads files from the kit. Check once: > `ls ${CLAUDE_SKILL_DIR}/../../../tokens >/dev/null 2>&1 && echo KIT_OK || echo KIT_MISSING` > On `KIT_MISSING` only the skill folders were installed, which is what > `npx skills add` does. Say so plainly, point the user at > `npx ux-ui-agent-skills init` or the plugin install, and stop. Do not guess the > contents of a file you could not open.
Generate a new design-project skeleton from `${CLAUDE_SKILL_DIR}/../../../templates/product-design/` into a target directory the user names.
The layout to produce (exactly the reference structure):
<target>/
# CONTEXT CLAUDE LOADS
CLAUDE.md the brief Claude reads every session
CLAUDE.local.md your personal prefs, gitignored
.mcp.json Figma, Notion, Drive connections
# TEAM TOOLKIT
.claude/
rules/ conventions, loaded only when relevant
skills/ repeatable workflows, off main context
commands/ your custom slash commands
settings.json shared permissions, checked into git
# YOUR PROJECT
design-tokens.json source of truth: color, type, spacing
src/components/ the real UI Claude reads and edits
public/images/ real images so prototypes don't break
reference/ real screens Claude studies for contextThe fast path, when the kit is installed or reachable via npx:
npx ux-ui-agent-skills new <target>
That does steps 2 and 3 below in one call (template copied, `CLAUDE.local.md` un-suffixed, engine areas installed without overwriting the lean brief). Then pick up at step 4. The manual steps are spelled out so the same result can be produced by hand from a clone.
Steps: 1. Ask for the target directory if not given. Create it if it does not exist. 2. Copy every file of `${CLAUDE_SKILL_DIR}/../../../templates/product-design/` into it, preserving the tree. Two rules on the copy:
only so the kit's own gitignore cannot swallow the file).
`src/components/`, `public/images/`, and `reference/` so the empty folders survive the first commit. The template already ships `.gitignore` (which ignores `CLAUDE.local.md`), `.mcp.json`, `.claude/settings.json`, and three rules files (`components.md`, `tokens.md`, `accessibility.md`). 3. Install the engine next to it so the gates the brief references actually run:
npx ux-ui-agent-skills add tokens components taste accessibility \
workflows content frameworks design-systems scripts skillsUse `add`, not `init`: `init` also writes the full engine `CLAUDE.md` over the lean project brief. The project brief stays short by design, and the engine knowledge loads from the copied folders when a task needs it. 4. Install the render gates' one dependency in the target: `npm i -D playwright`. Skipping this is silent: every render gate prints `SKIPPED` and exits 0. 5. Fill in the placeholders in `CLAUDE.md` with the user: product, primary user, stack, icon set, vocabulary. Leave nothing in angle brackets behind. 6. Point `design-tokens.json` at the brand: replace the `primitive.brand` ramp, keep the semantic and component tiers, then prove it still passes:
python3 ${CLAUDE_SKILL_DIR}/../../../scripts/validate_contrast.py7. Remind the user to fill `reference/` with real screens and `public/images/` with real imagery, and to set the MCP env vars (`FIGMA_API_KEY`, ...) in their own shell. Never write a secret into `.mcp.json`.
Verify before handing over: the target tree matches the layout above, and
python3 ${CLAUDE_SKILL_DIR}/../../../scripts/validate_template.pypasses against the source template.
Repo: plugin87/ux-ui-agent-skills
Audit a UI or design against WCAG 2.2 AA/AAA and ARIA patterns, returning…
Apply a visual direction — an archetype (high-end agency, editorial minimal, brutalist,…
Generate a complete, accessible brand design system from a brief — primitive → semantic →…
Adversarial design critique of the current work — render it, look at it, and argue for…
Build a dense, data-heavy screen - an analytics console or a trading terminal - with real…
Generate production-ready, accessible, token-driven component code for ANY framework —…