convez.ai
AgenteDesign & UIIntermediário
neylonjon

Jon Neylon Atomic Brand Kit

neylonjon/jonneylon-atomic-kit

Sistema atômico de marca pessoal executável por agentes IA

Ver no GitHub
InicialPluginScore · de 0 a 100

Kit de marca pessoal codificado em formato legível por agentes IA, contendo camadas verbal (voz, posicionamento, messaging) e visual (cores, tipografia, componentes, motion). Pronto para ser integrado em ferramentas como Claude Code, Cursor e Codex para gerar interfaces on-brand automaticamente.

Instalação

$npx convez-ai neylonjon/jonneylon-atomic-kit

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
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 2 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Gerar interfaces web mantendo consistência de marca pessoal
  • Treinar agentes IA com diretrizes de design e voz específicas
  • Acelerar prototipagem de landing pages e projetos com identidade visual coerente

Veredito

Implementação pragmática de design system como instrução para agentes, não documentação. Abordagem inovadora que codifica juízos criativos upstream, mas requer compreensão clara do que um agente pode vs. não pode gerar automaticamente (ver `magic_trick.md`).

README

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

Jon Neylon — Atomic Brand Kit

A machine-readable encoding of my personal brand. Drop it into Claude Code, Cursor, Codex, or any agent that reads Markdown / YAML / CSS / SVG and it should produce surfaces that feel like me — voice, type, color, motion, posture.

This is the brand as instructions, not documentation.

jonneylon-atomic-kit/
├── readme.md              ← you are here
├── magic_trick.md         ← what the kit cannot generate (read it)
├── human/                 ← human-readable references (PDFs, links)
└── agent/                 ← agent-readable instructions
    ├── verbal/            ← positioning, audience, voice, concepts
    └── visual/            ← tokens, type, color, motion, components

What this is

I'm Jon Neylon. Founding Design Engineer at NextWork, based in Austin. I design + code + build with AI agents. This kit is the same atomic-system thinking Little Plains ships to founders, applied to my personal brand instead of a startup's.

The thesis: when execution is increasingly agentic, value moves upstream to the encoded thinking. The kit is the deliverable; everything an agent produces from it is downstream.

Quickstart

git clone https://github.com/neylonjon/jonneylon-atomic-kit
cd jonneylon-atomic-kit

Then in your agent of choice (Claude Code, Cursor, Codex, etc.):

Read every file under agent/. Then build me a single-page HTML landing page that introduces a new project of mine called [name]. Use the verbal layer for voice, the visual layer for type/color/motion/components.

A correctness floor — voice, palette, type, layout rhythm — should come back automatically. Originality is magic_trick.md's job.

What's in agent/

verbal/ — the upstream judgment calls.

  • positioning.md — who I am, what I do, why
  • audience.yaml — who the work is for, in order
  • messaging.md — repeatable phrases, taglines, headline patterns
  • voice.md — how the words behave (no em-dashes, source-case, etc.)
  • concepts.md — the core ideas the brand recombines
  • differentiation.md — what I do that the median design engineer doesn't

visual/ — the executable layer.

  • colors_and_type.css — palette + type scale as CSS custom properties
  • fonts/ — Gosha Sans OTF/TTF (display + body)
  • tokens/ — palette, type, spacing as portable token files
  • motion/ — ease curves + reveal sequence + Paper Shaders config (CSS + JSON)
  • assets/ — JN mark (SVG), halftone shader, favicon
  • components/ — header, tile, button, hairline-sweep as portable HTML
  • artifacts/web/ — sample landing pages built from this kit

What's NOT in agent/

magic_trick.md. Read it before you ship anything from this kit. The agent layer gets you to a correct, branded floor. The magic trick is what tilts the output away from the median — and by Little Plains' framing, the part that can't be encoded. Read it. Don't skip it.

License

MIT for the structure and code. The verbal layer (positioning, voice, magic-trick reasoning) is mine — fork freely, but please don't ship it under your own name.

— Jon · jonneylon.com · @neylonjon

Ver o original no GitHub

Bom para

Relacionados