Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
neonwatty avatar

Looper

  • Updated May 3, 2026
  • neonwatty/claude-plugins

looper is a Claude Code skill in the Design & UI/UX category. Autonomous feature implementation with independent verification and design quality enforcement

Key points

  • looper
  • Design & UI/UX
  • AI-coding skill

Looper by the numbers

  • Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add neonwatty/claude-plugins
/plugin install looper@neonwatty-plugins

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Last updatedMay 3, 2026
Repositoryneonwatty/claude-plugins

What it does

Autonomous feature implementation with independent verification and design quality enforcement

README.md

Looper

Autonomous feature implementation with independent verification.

/implement {feature}
        │
        ▼
   ┌─────────┐
   │ Planner │ → Creates PLAN.md with 2-4 steps
   └────┬────┘
        │
   ┌────▼────┐      ┌─────────┐
   │ Worker  │ ───► │ Checker │  (independent sub-agent)
   │ (main)  │ ◄─── │         │
   └────┬────┘      └─────────┘
        │                │
        │          Uses Chrome MCP for
        │          visual verification
        │
   (if 3 failures)
        ▼
   Escalate to user

Installation

Option 1: Add Marketplace to Your Project

In your project's .claude/settings.json:

{
  "extraKnownMarketplaces": {
    "neonwatty": {
      "source": {
        "source": "github",
        "repo": "neonwatty/claude-plugins"
      }
    }
  },
  "enabledPlugins": [
    "looper@neonwatty"
  ]
}

Option 2: Install Globally

/plugin marketplace add neonwatty/claude-plugins
/plugin install looper@neonwatty

Commands

Command Purpose
/implement {feature} Full autonomous implementation with verification loop
/plan {feature} Create implementation plan only (no coding)
/check [step] Manually verify a step's implementation
/resume [feature] Resume in-progress implementation

How It Works

  1. Plan - Breaks feature into 2-4 verifiable steps
  2. Implement - Builds each step with distinctive design choices
  3. Check - Independent agent verifies via code review + Chrome MCP visual checks + design quality
  4. Loop - If issues found, fix and re-check (max 3 attempts)
  5. Escalate - After 3 failures, asks user for guidance
  6. Ship - Creates PR when all steps pass

Design Aesthetics

Looper enforces distinctive design choices to avoid "distributional convergence" (generic defaults). The visual checker will reject UIs that use:

Dimension Rejected Expected
Typography Inter, Roboto, system fonts Distinctive typefaces, extreme weight contrasts
Color Purple gradients, corporate blue Cohesive themes, cultural aesthetics
Motion No animations Staggered reveals, micro-interactions
Backgrounds Solid white/gray Gradients, patterns, layered depth

See guides/design-aesthetics.md for detailed guidance and code examples.

Project Configuration (Minimal)

Optionally create CLAUDE.md with just the app URL:

## Looper Configuration
- **App URL:** http://localhost:5173

That's it. Looper automatically:

  • Creates a Makefile with standard targets (make ci, make test-e2e, etc.)
  • Installs required dev dependencies (eslint, knip, playwright, vitest)
  • Creates GitHub Actions CI workflow

Plans Storage

Plans are stored in .plans/{feature-name}/PLAN.md and preserved after completion for reference.

Requirements

  • Chrome MCP server (for visual verification)
  • GitHub CLI (gh) for PR creation
  • Git repository

Updating

/plugin update looper@neonwatty

License

MIT

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.