convez.ai
SkillDesign & UIIntermediário
siddpokuri

IntelliFront

siddpokuri/Intellifront

IA para diagnóstico, crítica e redesign de frontend com componentes prontos

Ver no GitHub
InicialPluginScore · de 0 a 100

Sistema de inteligência frontend multi-agente que avalia, critica e redesenha websites com código HTML/CSS produção-ready. Integra-se com Claude Code, Cursor e v0 para transformar sites existentes em designs profissionais com componentes animados, interações 3D e estratégia de conversão.

Instalação

$npx convez-ai siddpokuri/Intellifront

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
6
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Avaliar e melhorar design visual de sites existentes
  • Detectar padrões de IA e templates genéricos em designs
  • Gerar código HTML/CSS com componentes animados e interações 3D
  • Otimizar layouts para conversão em sites de negócio

Veredito

Skill robusto e bem estruturado para designers e desenvolvedores que precisam de feedback crítico real e redesigns com produção imediata. Diferencia-se por detectar 'slop' e fornecer referências vivas do mercado, não apenas gerar código genérico — ideal para agências e produtos DTC.

README

A tradução para o português deste README ainda está na fila.

IntelliFront

Frontend intelligence system for evaluating, critiquing, and redesigning websites.

IntelliFront is a frontend intelligence system that works with Claude Code, Cursor, v0, Lovable, and any tool that supports SKILL.md files. It operates like a senior design agency — diagnosing problems, detecting AI-generated patterns, fetching live references from top sites in your niche, and generating production-quality redesign code with animated components, 3D interactions, and exact asset prompts.

Live Demos

Sites built entirely using IntelliFront + Claude Code — no manual design work.

  • KOVA — DTC sneaker brand
  • Seve — Luxury fragrance brand
  • Lumen — SaaS AI research tool

What It Does

  • Diagnoses visual hierarchy, typography, spacing, accessibility, and trust issues
  • Detects AI-generated, template-built, and low-credibility design patterns
  • Researches best-in-class examples live from the web — fetches real reference sites per niche
  • Redesigns sections with working HTML/CSS code, including animated components
  • Optimizes for conversion on business-oriented sites
  • Transforms entire sites in Godmode with a full redesign strategy
  • Asks upfront whether you want AI image prompts, stock photo search terms, or your own assets
  • Asks upfront how much animation you want (subtle / expressive / cinematic)
  • Tells you exactly what images, assets, and 3D components you need — with inline comments in the code

What It Is Not

IntelliFront is not a website generator. It does not create websites from nothing. It transforms existing or described sites into better ones.


Repository Structure

intellifront/
├── SKILL.md                          ← Entry point — load this first
│
├── agents/
│   ├── 01-critic.md                  ← Frontend Critic
│   ├── 02-design-intelligence.md     ← Design Intelligence + code generation
│   ├── 03-slop-detector.md           ← AI/template pattern detection
│   ├── 04-conversion.md              ← Conversion optimization
│   └── 05-godmode.md                 ← Full transformation mode
│
├── styles/
│   ├── design-system.md              ← Design principles, tokens, absolute bans
│   ├── typography.md                 ← Font selection protocol + pairing library
│   └── motion.md                     ← Motion tier system (Tier 1–5) + 3D patterns
│
├── components/
│   ├── README.md                     ← Component library overview
│   ├── nav.html                      ← Navigation variants
│   ├── hero.html                     ← Hero variants (split, editorial, kinetic)
│   ├── cards.html                    ← Feature strips, bento grid, testimonials
│   ├── buttons.html                  ← 15 button variants with decision guide
│   ├── interactive.html              ← Toggles, checkboxes, radios, loaders,
│   │                                    tab bars, badges, step indicators (22 components)
│   ├── cta.html                      ← CTA section variants
│   ├── footer.html                   ← Footer variants (minimal, structured, editorial)
│   └── ui-kit.html                   ← Extended micro-components
│
├── docs/
│   ├── design-references.md          ← Curated niche inspiration library
│   └── anti-slop-index.md            ← Full slop pattern taxonomy
│
├── utils/
│   ├── scoring.md                    ← Score calculation formulas
│   └── confidence.md                 ← Evidence discipline protocol
│
└── examples/
    └── example-output-saas.md        ← Complete example run (SaaS landing page)

Compatibility

IntelliFront works with any AI coding tool that supports SKILL.md files.

ToolHow to install
Claude Codeclaude mcp add or drop SKILL.md into project
CursorAdd SKILL.md to .cursorrules or project context
WindsurfAdd SKILL.md to project context
v0Paste SKILL.md as system prompt
LovablePaste SKILL.md as system prompt
Gemini CLIPass SKILL.md as context
OpenCodePass SKILL.md as context
Any agentPaste SKILL.md as system prompt

Installing

Option 1 — .skill file (Claude Code / compatible tools)

Download intellifront.skill from the releases and install directly.

Option 2 — Manual (any tool)

  1. Clone or download this repo
  2. Point your tool at SKILL.md as the entry point
  3. Make sure the full repo is in context — agents reference files in /agents, /styles, /components, and /docs

Option 3 — System prompt

Paste the contents of SKILL.md into your system prompt. For best results, include the other files in your context window too.


How to Use It

Basic usage

Share a site in any of these forms:

  • A URL: https://yoursite.com
  • A screenshot (paste or upload)
  • HTML/CSS source code
  • A description: "a SaaS landing page for a project management tool called X"

IntelliFront will analyze the input, present agent options, and ask two quick questions:

  • Images — AI prompts, stock search terms, or you'll supply your own
  • Animations — subtle / expressive / cinematic

Selecting agents

Reply with your mode + image preference + animation tier, e.g:

  • godmode, a, 2 — Godmode, AI image prompts, expressive animations
  • 1+3, b, 1 — Critique + Slop Check, stock search terms, subtle animations
  • your call — IntelliFront decides everything based on the brand

Example prompts

Roast my landing page: [paste code or URL]
Does this look AI-generated? [paste code or screenshot]
Run godmode on this. I want AI image prompts and cinematic animations.
Run agents 1 and 3 on this: [paste code]
What's wrong with my hero section? [paste hero HTML]

The Agent System

AgentWhat it doesGenerates code?
1 — Frontend CriticDiagnoses problemsNo
2 — Design IntelligenceResearch + redesign directionYes
3 — Slop DetectorAI/template pattern detectionYes (for fixes)
4 — Conversion OptimizerCTA, funnel, copy improvementsYes (copy)
5 — GodmodeFull transformation: all agents + complete redesignYes (full)

Generated Code

When agents generate code, it is:

  • Self-contained — HTML with inline <style> blocks, no framework dependencies
  • Mobile-first — all layouts work at 375px viewport
  • Accessible — semantic HTML, aria labels, focus states
  • Original — written for the specific site, not copied from components verbatim
  • Adaptable — generated code can optionally use the IntelliFront component patterns from /components/, but this is not required

Code style and design system usage is flexible — agents write code appropriate to the specific project, not forced into a template.


Design Philosophy

Restraint over maximalism. The most common failure in AI-generated design is doing too much. IntelliFront pushes in the opposite direction — every element earns its place.

Motion is tiered, not blanket. A legal firm and a creative agency have different emotional contracts. Animations scale from "opacity only" (Tier 1) to "cinematic and physics-based" (Tier 5) based on brand fit — not on what's impressive to generate.

Evidence before advice. Every finding is labeled FACT, INFERENCE, or ASSUMPTION. Fabricated metrics, invented competitor comparisons, and hallucinated user behavior are explicitly forbidden by the confidence protocol in utils/confidence.md.

Slop is specific, not vague. "This looks AI-generated" is not a useful critique. "This uses the purple-to-blue gradient hero (V1), Inter as a display face (T1), and the transformation-verb headline (C1)" is actionable.


Niche Coverage

IntelliFront's design reference library covers:

  • SaaS / developer tools / AI products
  • Fintech / financial services
  • E-commerce / DTC brands
  • Real estate / luxury property
  • Agencies / creative studios
  • Healthcare / wellness
  • Restaurants / hospitality
  • Personal portfolios

Reference sites and pattern synthesis in docs/design-references.md.


Contributing

This skill is designed to evolve. High-value contributions:

  • New slop patterns — add to docs/anti-slop-index.md with ID, description, detection signal, and fix direction
  • New component variants — add to the relevant components/*.html file as a commented variant
  • New niche references — add to docs/design-references.md under the appropriate niche
  • New font pairings — add to styles/typography.md under the relevant brand personality
  • Example outputs — add real (anonymized) IntelliFront runs to examples/

License

MIT. Use it, adapt it, build on it.


IntelliFront is a Claude skill — it runs inside Claude, not as a standalone application. It requires access to Claude (claude.ai or Anthropic API).

Ver o original no GitHub

Bom para

Relacionados