How to use the Anti-UI-Slop Design Guide agent skill

What the skill makes the agent do

The skill is aimed at agents that ship generic UI. Before writing code, the agent is asked to identify the screen's real job, its primary user and action, and the content it must show. It also covers required loading, empty, error, success, disabled and permission states.

The agent is told to reuse the repository's existing components, semantic tokens, typography, spacing and interaction conventions before adding anything new. It should use product-specific labels and data, and not invent metrics, activity, testimonials or users to make a layout look complete.

• For a new interface or major redesign, the agent writes a short design contract covering hierarchy, workflow, allowed components, required states, responsive behaviour and acceptance criteria.
• Smaller changes stay small.
• When the environment allows it, the agent renders the result once and fixes visible problems such as clipping, overlap, distorted media and controls that do not work.
• The agent runs the project's normal checks and keeps its handoff concise.

Install the anti-ui-slop skill in Claude

Skills are folders of instructions that an agent loads when a task matches them. The skill is distributed as a ZIP file from the agent-skills directory page. Unzip it into your personal skills folder at ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it to apply only to that project.

If you use the Claude apps instead, you can upload the skill under Settings, then Capabilities, then Skills.

The skill's own source also gives a direct install command from its canonical source. It is shown below as written there.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project folder
• Claude apps: Settings, then Capabilities, then Skills, then upload

What you need before you start

The skill lists no required software or dependencies. The free skill and the public UIZZE screenshot catalogue work without an account, token, dependency, script or executable.

The skill works best when your project already has a design system, a product brief or a clear description of the screen. The source is direct on this point: without those, the output is more likely to be generic. The skill is built to extend an existing design system rather than replace one.

• No account is needed for the free skill or the public catalogue.
• No API token is needed for the basic skill.
• Best results come from a project with existing components and design tokens.
• Best results also come from a short written brief stating the screen's purpose and user.

How to trigger the skill with natural requests

You do not need special syntax. Describe the interface task in plain language and mention the screen, the user and the states you care about. The skill is meant for designing, implementing, redesigning, critiquing and doing a pre-ship review of web or iOS interfaces.

Name the states you need, because the skill checks for them. If you want a design review rather than new code, say so, since the skill covers critiquing as well as building.

• Build a new account settings screen for our existing design system, covering loading, empty and error states.
• Redesign the billing page in our app. Write a short design contract first with acceptance criteria.
• Review this onboarding flow before we ship. Check for missing states and layout problems.
• Add a team invite screen using our existing components only, and render it to check for overlap.

Optional UIZZE visual references

The skill can optionally consult UIZZE's public screenshot catalogue. It can also use an authenticated UIZZE search tool, described in the source as an MCP exposing find_ui_references and find_ui_materials. The agent uses these only when those tools are actually available in your host, and only when a concrete visual question would benefit from evidence.

If a search returns nothing, the agent continues from your repository. References are treated as inspiration. The skill says to take decisions about hierarchy, density, navigation and controls, and never to copy another product's branding, text, imagery or exact layout.

• Use references for a specific question, such as how dense a settings list should be.
• Expect that some searches will return nothing useful.
• Do not treat any reference as permission to copy another product's identity or assets.

Safety notes: scripts and network access

The skill's listing states that it runs no scripts. It does use the network. The public catalogue and the optional UIZZE tools are online services, and the install command fetches the skill from uizze.com.

Review the skill's files before installing if your policies require it. The source file listed is SKILL.md. The skill is licensed under MIT, and the source names its license file in the repository.

• Runs scripts: no
• Uses network: yes, for the optional catalogue and UIZZE tools
• Only SKILL.md is listed among the files