Gemini Designer
oil-oil/gemini-designer
Consultor de design visual externo para agentes Claude via Gemini
Ver no GitHubSkill que delega tarefas de design visual para o Gemini via OpenRouter, oferecendo crítica de UI, mockups HTML, ícones SVG e feedback visual baseado em arquivos. Funciona como um consultor de design externo que agentes podem chamar para obter julgamento visual sem modificar diretamente os arquivos do projeto.
Instalação
npx convez-ai oil-oil/gemini-designer1 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
Casos de uso
- Obter crítica de UI e direção visual de designs existentes
- Gerar mockups HTML completos baseados em briefings
- Criar ícones SVG e wordmarks a partir de descrições
- Analisar screenshots e fornecer feedback visual estruturado
- Estabelecer imagética e direção de design para produtos
Veredito

Skill bem estruturado que integra o Gemini como consultor de design externo. Ideal para agentes que precisam de feedback visual sem dependência exclusiva de Claude, mas exige API key do Gemini e adota padrão CLI limpo que respeita autonomia do agente sobre aplicação de mudanças.
README
A tradução para o português deste README ainda está na fila.
Gemini Designer
Gemini Designer is a visual-design advisor skill for agents. It helps with design direction, UI critique, HTML mockups, SVG icons, handwritten wordmarks, and file-based visual feedback.
The skill asks Gemini for visual judgment through the bundled gemini-designer CLI, then lets the main agent decide how to apply the advice in the current workspace.
Install
The easiest way is to give this GitHub repository to an agent such as Codex, Claude Code, or Cursor and ask it to install the skill:
https://github.com/oil-oil/gemini-designer
You can also install it directly from a terminal:
npx skills add oil-oil/gemini-designer
After installation, agents should use the gemini-designer skill when a task needs external visual design judgment.
Authorization
The CLI reads local configuration from:
~/.config/gemini-designer/config.toml
By default, the API key is read from:
~/.config/gemini-designer/api_key
Do not commit API keys or local config files. The repository ignores common local secret files, including .env, config.toml, and api_key.
What Agents Should Know
- Gemini is stateless. It only sees the current prompt, files passed with
-f, and images passed with-i. - For existing UI, use
gemini-designer advise. - For broad art direction or design imagery markdown, use
gemini-designer direction. - For new standalone HTML mockups, use
gemini-designer html. - For SVG icons, simple illustrations, and single handwritten wordmarks, use
gemini-designer svg. - Pass complete relevant files when Gemini needs to judge an existing design.
- Pass screenshots or visual references with
-iwhen the visible result matters. - Do not ask Gemini to patch project files directly. Use its advice, then apply the changes in the workspace.
CLI
The skill installs a global command:
gemini-designer
Typical examples:
gemini-designer advise "给这个页面提视觉设计建议" -f ./design.html -o design-page-advice.md
gemini-designer direction "给这个产品生成设计意象 markdown" -o product-design-imagery.md
gemini-designer html "生成一个完整的产品页面设计稿" -f ./brief.md -o ./designs/product-page.html
gemini-designer svg "为 Museon 生成一个手写 SVG 字标" -o museon-wordmark.svg
Bare output filenames are saved under .gemini-designer/ in the current workspace.
Repository Layout
SKILL.md
scripts/gemini-designer
scripts/install_cli
SKILL.md tells agents when and how to use Gemini. scripts/install_cli installs the CLI into the user's local bin directory. scripts/gemini-designer is the command agents call.