UIZZE · Better UI for Coding Agents

A focused UI design workflow for Codex, Claude Code, Cursor and Copilot, with optional references from 800,000+ real screens.

Source repo
uizze/uizze
Stars
★ 121
Last updated
21d ago
License
MIT
Primary language
JavaScript

At a glance

How it runs
Agent plugin / skillCLIMCP server
Works with
Portable with changesCodex · Claude CodeChatGPT (Partial support)
Cost
Free tier plus a paid hosted plan
Setup effort
Low · running in minutes
You'll need
Node.js (npx)Shell / CLINetwork accessLocal filesystemMCP Server
Typical use
Redesigning a billing settings page in an existing SaaS codebase so the current plan, payment method and invoices scan quickly, with loading, empty, error and permission states covered by ui-design.
Not a fit if
  • Teams that need fully offline operation with no reference lookup
  • Teams with zero budget that still need live UI reference search
  • Backend-only projects with no frontend interface work
Website
uizze.com ↗ set on GitHub
Maintenance
Commits in the last 30 dayslast commit 2026-10-31latest release v1.3.1 (2026-10-03)+3.1 ★/day over 7 days
Source review
69/100 · Some gaps

What does this agent do, and when should you use it?

UIZZE is a repository of UI design skills and checks aimed at coding agents rather than at hand-written design work. Its free core is three installable skills: ui-design for building or redesigning a screen, anti-ui-slop for fixing a generic first draft, and ui-radar for investigating a specific UI decision against references. Skills install through npx skills add from the uizze.com domain packaging, with uizze/uizze on GitHub mirroring the same three skills, and they work without an account. An optional paid MCP server adds two tools, find_ui_references and find_ui_materials, that retrieve up to three full-screen web or iOS references and up to three fonts, icons, animated icons or requested material packs per call. The repo also ships a free UI Slop Gate GitHub Action that scans changed frontend source in pull requests for inert controls, missing state markers, hardcoded colors and combinations of generic dashboard cues. It does not generate your interface for you: agents still implement screens using your own components and design tokens.

The repository contains three skill packages under skills/ui-design, skills/anti-ui-slop and skills/ui-radar, installed with npx skills add https://uizze.com --skill <name> or, from the GitHub mirror, npx skills add uizze/uizze --skill ui-design. Once installed, you hand the agent a real task as a prompt; the skill guides it through information hierarchy, reuse of existing components and design tokens, coverage of loading, empty, error and permission states, and inspection of the result at desktop and mobile sizes. With the optional paid MCP connected, the agent can call find_ui_references for up to three full-screen references per result and find_ui_materials for up to three hosted fonts, icons, animated icons or explicitly requested packs; when a search returns no useful evidence, work continues from the project. Separately, the UI Slop Gate Action (uses: uizze/uizze@v1) reads changed frontend source in a pull request, reports findings inline and in the workflow summary, and can fail the run with fail-on: error. Supporting distribution includes a Next.js starter, a Storybook integration, and Gemini CLI and Antigravity plugins.

  1. Redesigning a billing settings page in an existing SaaS codebase so the current plan, payment method and invoices scan quickly, with loading, empty, error and permission states covered by ui-design.
  2. A landing page just came out of an agent looking like a template; install anti-ui-slop to push product-specific content, deliberate layout and a finish review.
  3. You are about to build a team permissions screen with inherited roles and restricted actions and want to see how real products handled it, so you connect the MCP first and implement second.
  4. You want to catch inert buttons and hardcoded colors before merge, by adding the UI Slop Gate Action to a frontend repository without any account or API key.
  5. You are trying your first screen in Claude Code, Cursor or GitHub Copilot and want a guided path, using the billing-screen walkthrough to verify the skill actually changes output.
  6. You are designing an iOS onboarding and account-recovery flow and want the native SwiftUI starter and browser prototype the repo points to.

Key setup facts

Pulled from the install and usage notes below. The project README has the current values.

Install / run
  • npx skills add https://uizze.com --skill ui-design
  • npx skills add https://uizze.com --skill anti-ui-slop
  • npx skills add uizze/uizze --skill ui-design

How do you install or deploy this agent?

Install the free UI design skill (no account needed; skills work without an MCP connection):

npx skills add https://uizze.com --skill ui-design

For a focused anti-slop review instead:

npx skills add https://uizze.com --skill anti-ui-slop

Prefer installing from GitHub? Use the mirror path:

npx skills add uizze/uizze --skill ui-design

To add the pull-request gate, put this in your workflow:

permissions:
  contents: read

steps:
  - uses: actions/checkout@v7
    with:
      fetch-depth: 2
      persist-credentials: false
  - uses: uizze/uizze@v1
    with:
      fail-on: error

The optional paid MCP requires creating an agent token in UIZZE first, then following the MCP setup guide under integrations/mcp.

How do you use this agent?

After installing ui-design, give your agent a real task and tell it to reuse your components and tokens:

Use ui-design to improve our billing settings page. Make the current plan,
payment method, and invoices easy to scan. Reuse our components and design
tokens. Include loading, empty, error, and permission states. Inspect the
result at desktop and mobile sizes and fix what breaks.

With the MCP connected, ask for evidence before implementation:

Find references for a team permissions screen with inherited roles and
restricted actions. Explain which decisions would help our workflow,
then implement them using our existing components and visual language.

To reproduce the pull-request check locally, follow the source, findings and local command documented in the pull-request-check example.

What are this agent's strengths and limitations?

Pros
  • The free skills need no account and no MCP connection, so the only adoption cost is a single npx install command.
  • It covers three distinct stages rather than one prompt: building interfaces, fixing generic drafts, and investigating UI decisions.
  • The UI Slop Gate runs as a GitHub Action with no account, API key or source upload, reporting findings next to the code and in the workflow summary.
  • Documented environment coverage is broad: Codex, Claude Code, Cursor and GitHub Copilot, plus Cline, Gemini CLI, Antigravity, VS Code, Windsurf and Zed as MCP clients.
Limitations
  • Live reference and material retrieval depends on the optional paid MCP, so the free path cannot query the 800,000+ screen library on demand.
  • The skills do not produce finished interfaces; results depend on whether your project already has components and design tokens, and on the model you use.
  • The UI Slop Gate is a deliberately conservative source check, and the README asks you to keep rendered inspection and normal accessibility and usability testing alongside it.
  • Bundled skill playbooks keep an Apache-2.0 license and third-party notices that differ from the repository's MIT terms, so redistributing a skill package requires reading LICENSING.md first.

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

Agent Source review Form / cost Stars Updated Language Full support on
UIZZE · Better UI for Coding Agents This agent 69 · Some gaps Agent plugin / skillFreemium ★ 121 21d ago JavaScript Codex · Claude Code
Open Design 70 · Some gaps Desktop appFree + model costs ★ 109k 1d ago TypeScript Codex · Claude Code · OpenAI API · Claude API
Dembrandt UX Skills 60 · Some gaps Agent plugin / skillFree ★ 197 1d ago JavaScript Claude Code
Superdesign Skill 56 · Major gaps Agent plugin / skillFreemium ★ 1k 3mo ago JavaScript Codex · Claude Code

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Some gaps
69/ 100 5-point scale 3.5 / 5
Trust 18/29
Reliability 6/14
Adaptability 16/18
Convention 13/18
Effectiveness 10/13
Verifiability 6/8
Why each dimension lost points
Trust18 / 29 · 3.1/5

The skills explicitly prioritize user requests and project constraints, require inspecting project files, and describe optional MCP authentication and the need not to claim an unverified connection. ui-radar treats screenshots and retrieved content as untrusted. The GitHub Action confines paths to the workspace and rejects traversal and external symlinks; CI permissions are read-only. Dependency security and rollback guidance are limited: the sample shows no general dependency audit or undo procedure, so these do not earn full marks.

Reliability6 / 14 · 2.1/5

Skill routing and responsibilities are fairly clear, and CI checks skill names, checksums, and some integrations. However, README claims broad agent compatibility while actual capabilities depend on host support for file access, rendering, or external connections; version fields in the sample are also inconsistent. Skills ask the agent to disclose service errors and unavailable rendering, and the action emits error annotations, but not all failure paths are shown.

Adaptability16 / 18 · 4.4/5

The skills cover new work, redesign, polish, simplification, audits, native iOS, and screenshot recreation, with instructions to select relevant playbooks. They define optional MCP access, make Overdrive opt-in, and set boundaries for missing connections, unreadable images, and unavailable rendering. README and metadata support broad host compatibility, though actual fit depends on host tools.

Convention13 / 18 · 3.6/5

The README provides a skill directory, installation options, MCP setup, and examples; skill structure and licensing notices are clear. The repository contains both MIT and Apache-2.0 material and points to a license map. Plugin metadata says 1.3.1 while sampled skills use differing version markers, and the packet shows no changelog or clear maintenance owner, limiting naming and maintenance scores.

Effectiveness10 / 13 · 3.8/5

The workflows connect product constraints, relevant references, implementation, render checks, and concise handoff. ui-radar limits the number of references and ties visible observations to design decisions, making its output actionable. Value depends on task fit and host capabilities; paid MCP access is useful only for some retrieval tasks, so the evidence does not establish that its cost is worthwhile in every case.

Verifiability6 / 8 · 3.8/5

Core workflows and constraints are traceable to specific skills, with partial corroboration from the README, plugin metadata, and CI configuration for installation and repository contracts. Skills require distinguishing observations, recommendations, service errors, and unverified results. Promotional screen-count claims are not independently supported in this sample, and version markers are not fully consistent across files.

Risks and how to mitigate them
  • This review covers only the supplied file sample; unseen files are unknown, not proven absent or verified.
  • Promotional screen-count claims are not independently substantiated in the sample, and host-dependent rendering and MCP capabilities were not executed or verified in this static review.
Evidence confidence: Low Reviewed Nov 07, 2026 Reviewed revision fc401e9e654c
See the full review method →
Maintain this project? Add the FARS badge to your README

The badge shows the current review score and verdict and links back to this page. It refreshes about once an hour after a re-review.

FARS

[![FARS review](https://followagents.com/badge/uizze.svg)](https://followagents.com/en/agents/uizze)
<a href="https://followagents.com/en/agents/uizze"><img src="https://followagents.com/badge/uizze.svg" alt="FARS review"></a>

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

  1. v1.3.1 v1.3.1: source-only packaging and provenance patch
    Release notes excerpt
    Source-only patch from the reviewed, exact-head PR #345. Native plugin/catalog package versions are now 1.3.1, with the existing feature lineage, branding, prices, GitHub Action pin and authenticated MCP configuration preserved. The anti-ui-slop and ui-design packages retain their Apache-2.0 licenses, third-party MIT notices, playbooks and checksums. The ui-radar source package declares MIT and retains the complete existing copyright and permission grant both inline in its standalone Markdown and in its adjacent LICENSE. Tool metadata is narrowly scoped to Read, Glob, Grep and WebFetch. Actual …
    View on GitHub ↗
  2. v1.3.0 Uizze 1.3.0 — Design craft and opt-in Overdrive
    Release notes excerpt
    UI design that stands out. The free anti-ui-slop and ui-design skills now support direction selection, contextual craft checks, honest reference-result reporting, and visual re-verification. Optional Overdrive helps with explicitly requested interaction work: ask “Use Uizze Overdrive on this screen.” A direction must be chosen before implementation; no native slash command or executable engine is added. …
    View on GitHub ↗
  3. storybook-v0.1.2 Storybook addon 0.1.2
    Release notes excerpt
    The Storybook addon now keeps clipboard feedback specific to the selected story and contract. If clipboard access fails, the panel provides selectable Markdown and a retry; delayed results from another story no longer overwrite current feedback. Clean package installation and browser checks passed with React/Vite on Storybook 9.1.20 and 10.6.0. CI now checks both versions. The development toolchain and local preset were repaired so declaration builds, typechecking and linting pass. Install: ```sh …
    View on GitHub ↗

FAQ

What do I get for free versus with the paid MCP?
The three skills — ui-design, anti-ui-slop and ui-radar — are free and need neither an account nor an MCP connection. The paid MCP adds the find_ui_references and find_ui_materials tools, letting the agent pull up to three full-screen references and up to three fonts, icons or packs per call; if a search returns no useful evidence, continue from the project.
Does the UI Slop Gate upload my source or need a key?
No. The Action reads changed frontend source directly and requires no account, API key or source upload. It looks for inert controls, missing state markers, hardcoded colors and combinations of generic dashboard cues.
Will this write the interface for me?
No. It supplies skills, optional reference retrieval and a pull-request gate; your coding agent still implements the screen using your project's own components and design tokens, so a reusable design system in the project is a practical prerequisite.
Can the UI Slop Gate replace accessibility and usability testing?
No. It is a conservative source-level check, and the README explicitly recommends using rendered inspection plus your normal accessibility and usability checks alongside it.
Can I redistribute the skills commercially?
Repository-authored code and text are MIT, but bundled skill playbooks retain Apache-2.0 licensing and third-party notices, so check LICENSING.md before redistributing a skill package.
Is UIZZE · Better UI for Coding Agents open source?
Yes. The source code is on GitHub (uizze/uizze) under the MIT licence.
Is UIZZE · Better UI for Coding Agents safe to use?
FollowAgents' FARS review (Nov 07, 2026) scored UIZZE · Better UI for Coding Agents 18/29 on trust and safety. None of the eight safety controls the review checks for was missing outright.
Is UIZZE · Better UI for Coding Agents still maintained?
Yes. The latest commit on GitHub is from 2026-10-31, within the last 30 days. The latest release is v1.3.1 (2026-10-03).
Who makes UIZZE · Better UI for Coding Agents?
UIZZE · Better UI for Coding Agents is published on GitHub by the account uizze. FollowAgents has not verified which person or company is behind that account.
What is UIZZE · Better UI for Coding Agents's official website?
The project's GitHub repository lists https://uizze.com as its website. That link is set by the repository owner; the source code itself is on GitHub (uizze/uizze).
Was this page helpful?
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related concepts