stitch-mcp-generate-screen-from-text
Generate high-fidelity UI screens or wireframes from text descriptions via Stitch MCP generate_screen_from_text. Use when the user wants to create, design, or visualize a UI screen using Stitch. Supports MOBILE, DESKTOP, TABLET, and SMART_WATCH device types with Gemini Pro or Flash models.
Best use case
stitch-mcp-generate-screen-from-text is best used when you need a repeatable AI agent workflow instead of a one-off prompt. It is especially useful for teams working in multi. Generate high-fidelity UI screens or wireframes from text descriptions via Stitch MCP generate_screen_from_text. Use when the user wants to create, design, or visualize a UI screen using Stitch. Supports MOBILE, DESKTOP, TABLET, and SMART_WATCH device types with Gemini Pro or Flash models.
Generate high-fidelity UI screens or wireframes from text descriptions via Stitch MCP generate_screen_from_text. Use when the user wants to create, design, or visualize a UI screen using Stitch. Supports MOBILE, DESKTOP, TABLET, and SMART_WATCH device types with Gemini Pro or Flash models.
Users should expect a more consistent workflow output, faster repeated execution, and less time spent rewriting prompts from scratch.
Practical example
Example input
Use the "stitch-mcp-generate-screen-from-text" skill to help with this workflow task. Context: Generate high-fidelity UI screens or wireframes from text descriptions via Stitch MCP generate_screen_from_text. Use when the user wants to create, design, or visualize a UI screen using Stitch. Supports MOBILE, DESKTOP, TABLET, and SMART_WATCH device types with Gemini Pro or Flash models.
Example output
A structured workflow result with clearer steps, more consistent formatting, and an output that is easier to reuse in the next run.
When to use this skill
- Use this skill when you want a reusable workflow rather than writing the same prompt again and again.
When not to use this skill
- Do not use this when you only need a one-off answer and do not need a reusable workflow.
- Do not use it if you cannot install or maintain the related files, repository context, or supporting tools.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/stitch-mcp-generate-screen-from-text/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How stitch-mcp-generate-screen-from-text Compares
| Feature / Agent | stitch-mcp-generate-screen-from-text | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Generate high-fidelity UI screens or wireframes from text descriptions via Stitch MCP generate_screen_from_text. Use when the user wants to create, design, or visualize a UI screen using Stitch. Supports MOBILE, DESKTOP, TABLET, and SMART_WATCH device types with Gemini Pro or Flash models.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
## Tools
This skill is designed to call the Stitch MCP tool:
* `generate_screen_from_text`
If your client namespaces MCP tools, it may appear as `mcp__<serverName>__generate_screen_from_text`.
## When to use this skill
**CRITICAL PREREQUISITE:**
**You must ONLY use this skill when the user EXPLICITLY mentions "Stitch".**
**ALWAYS use this skill when the user:**
- Describes a UI interface **and asks Stitch to generate it**.
- Asks to "Design", "Generate", "Create", or "Make" a screen **using Stitch**.
- Provides specific visual requirements ("Dark mode", "Blue button") for a Stitch generation.
- Wants to visualize a wireframe or concept **via Stitch**.
- Is in the **Step 5** of the `stitch-ui-designer` SOP workflow.
**Trigger phrases include:**
- "Use Stitch to design a screen" (用 Stitch 设计一个页面)
- "Stitch generate UI" (Stitch 生成 UI)
- "Draw a login page with Stitch" (用 Stitch 画一个登录页)
## Input Parameters
The skill expects you to extract the following information from the user request:
* **`projectId`** (required): The numeric Project ID. **Format**: Pure ID (e.g., `37803...`), **NO** `projects/` prefix.
* **`prompt`** (required): The structured text description of the screen (see "Constructing the Prompt" below).
* **`deviceType`** (optional): The target device.
* Values: `MOBILE` (default), `DESKTOP`, `TABLET`, `SMART_WATCH`.
* **`modelId`** (optional): The model to use.
* Values: `GEMINI_3_PRO` (Recommended for quality), `GEMINI_3_FLASH` (Speed).
## How to use this skill
### 0. Call the MCP Tool
Invoke `generate_screen_from_text` with:
* `projectId` (pure numeric string, no `projects/`)
* `prompt`
* `deviceType` (optional)
* `modelId` (optional)
### 1. Constructing the Prompt (The Art of Prompting)
The `prompt` argument is the most critical factor for quality. Do not just pass the user's raw input. You **MUST** enrich it using the **Structure Strategy**:
`[Device] [Mode] [Screen Type]. [Style]. [Layout]. [Components].`
* **Context**: "Mobile High-Fidelity login screen."
* **Style**: "Cyberpunk aesthetic. Dark mode. Neon blue accents."
* **Layout**: "Center-aligned vertical stack."
* **Components**: "Glitch-effect Logo. Input fields with glowing borders. Primary 'Jack In' button."
### 2. Choosing Device Type (`deviceType`)
* `MOBILE` (Default): Vertical layouts, ~375px width. Best for consumer apps.
* `DESKTOP`: Horizontal layouts, ~1440px width. Best for SaaS, Dashboards, Landing Pages.
* `TABLET`: Hybrid layouts.
* `SMART_WATCH`: Tiny, compact layouts.
### 3. Choosing Model (`modelId`)
* `GEMINI_3_PRO`: **Recommended**. High intelligence, better instruction following, superior aesthetics. Use for all complex/final designs.
* `GEMINI_3_FLASH`: Faster, lower cost. Good for simple wireframes or rapid iteration.
## Best Practices
1. **Detailed Components**: Don't just say "Form". Say "Form with Email, Password, and Eye toggle icon".
2. **Color Precision**: Mention specific colors (e.g., "Emerald Green", "#FF5733") if the user specifies them.
3. **Content Realism**: Ask for realistic text placeholders (e.g., "Welcome back, Alice" instead of "Lorem Ipsum").
4. **Device Alignment**: Ensure the `prompt` description matches the `deviceType` (e.g., don't ask for a "Sidebar" on `MOBILE`).
5. **No Code Generation**: This skill generates **Visual Designs**, not implementation code. Do not confuse with coding skills (like `uniappx-project-creator`).
## Output Handling
`generate_screen_from_text` returns session info (e.g., `sessionId` and `outputComponents`). It may not return a screenshot directly.
After the generation completes, retrieve the resulting screen(s) via:
1. `list_screens` with `projectId` in the format `projects/{id}`.
2. `get_screen` with the selected `screenId` to fetch screenshot / html assets.
## Keywords
**English keywords:**
generate screen, design ui, create interface, make page, draw wireframe, text to ui, ui generation, stitch gen, mobile design, desktop design, dashboard, login, prompt engineering
**Chinese keywords (中文关键词):**
生成页面, 设计UI, 创建界面, 画图, 制作网页, 文本生成UI, 界面设计, 移动端设计, 桌面端设计, 仪表盘, 登录页, 线框图, 生成代码
## References
- [Examples](examples/usage.md)
- [Desktop Dashboard Example](examples/desktop_dashboard.md)
- [Mobile App Example](examples/mobile_app.md)
- [Wireframe Example](examples/wireframe.md)Related Skills
threejs-textures
three.js textures: Texture, DataTexture, CubeTexture, CompressedTexture variants, DepthTexture, VideoTexture, CanvasTexture, 3D/array textures, Source; sampling parameters, mipmaps, anisotropy, wrap/mag/min filters; PMREMGenerator in Extras for environment map prefiltering. Use when configuring GPU texture objects and PMREM; for Draco/KTX2 transcoder file paths use threejs-loaders; for material map slots use threejs-materials; for output color pipeline use threejs-renderers.
tauri-app-splashscreen
Configure a splashscreen window in Tauri v2 to display during app initialization and avoid white screen flashes. Use when adding a loading screen at startup, controlling splashscreen-to-main-window transitions, or handling slow frontend loads.
tui-textarea
Generate and render a pixel-precise ASCII TUI Textarea component with complete output blocks (TUI_RENDER, COMPONENT_SPEC, PENCIL_SPEC, PENCIL_BATCH_DESIGN) for Pencil MCP drawing workflows. Use when the user asks to create a textarea in a terminal UI, text-based interface, or Pencil MCP project.
tui-text
Generate and render a pixel-precise ASCII TUI Text component with complete output blocks (TUI_RENDER, COMPONENT_SPEC, PENCIL_SPEC, PENCIL_BATCH_DESIGN) for Pencil MCP drawing workflows. Use when the user asks to create a text in a terminal UI, text-based interface, or Pencil MCP project.
tui-full-screen
Generate and render a pixel-precise ASCII TUI Full Screen component with complete output blocks (TUI_RENDER, COMPONENT_SPEC, PENCIL_SPEC, PENCIL_BATCH_DESIGN) for Pencil MCP drawing workflows. Use when the user asks to create a full screen in a terminal UI, text-based interface, or Pencil MCP project.
stitch-vue-vant-components
Convert Stitch designs into modular Vite + Vue 3 + Vant 4 mobile components. Use when the user mentions Vant or Vue Mobile conversion from Stitch. Retrieves screen HTML via Stitch MCP get_screen, maps Tailwind to Vant tokens (mobile-first 375px, safe area), enforces Vue SFC structure with Vant 4 components (van-button, van-field, van-nav-bar, van-tabbar).
stitch-vue-layui-components
Convert Stitch designs into modular Vite + Vue 3 + Layui-Vue components. Use when the user mentions Layui conversion from Stitch. Retrieves screen HTML via Stitch MCP get_screen, maps Tailwind to Layui tokens (2px/4px radius), enforces Vue SFC structure with Layui-Vue components (layui-btn, layui-input, lay-card, lay-page-header).
stitch-vue-element-components
Convert Stitch designs into modular Vite + Vue 3 + Element Plus components. Use when the user mentions Element Plus, element-ui, or Vue Desktop conversion from Stitch. Retrieves screen HTML via Stitch MCP get_screen, maps Tailwind to Element Plus tokens, enforces Vue SFC structure with el-* components (el-button, el-form, el-table, el-card, el-menu).
stitch-vue-bootstrap-components
Convert Stitch designs into modular Vite + Vue 3 + BootstrapVue/BootstrapVueNext components. Use when the user mentions Bootstrap or BootstrapVue conversion from Stitch. Retrieves screen HTML via Stitch MCP get_screen, maps Tailwind to Bootstrap utilities, enforces Vue SFC structure with Bootstrap components (b-container, b-row, b-col, b-button, b-card).
stitch-uviewpro-components
Convert Stitch designs into uni-app + Vue 3 + uView Pro pages and components. Use when the user mentions uView Pro, uviewpro, or uni-app Vue 3 conversion from Stitch. Retrieves screen HTML via Stitch MCP get_screen, maps Tailwind to rpx/theme, enforces u-* component contracts (u-tabs, u-form, u-picker, u-card) with script setup.
stitch-uview-components
Convert Stitch designs into uni-app + Vue 2 + uView 2.0 pages and components. Use when the user mentions uView, uView 2, or uni-app Vue 2 conversion from Stitch. Retrieves screen HTML via Stitch MCP get_screen, maps Tailwind to rpx/theme, enforces uni-app page structure with uView 2 u-* components (u-navbar, u-form, u-button, u-cell-group).
stitch-ui-prompt-architect
Build Stitch-ready prompts from vague UI ideas (Path A) or from a Design Spec plus user request (Path B). Use when polishing a UI prompt before Stitch generation, improving a prompt that produced poor results, or merging a Design Spec with framework contracts (uView, Element Plus, Layui, Bootstrap, Vant) into a sectioned Context/Layout/Components prompt.