跳到正文
cccskills
免费在 GitHub 公开

stitch-mcp-edit-screens

Edits existing Stitch screens with text prompts — the iteration tool. Change colors, layout, content, or style without regenerating from scratch. Supports refinement loops via output_components suggestions.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(1)

  • SKILL.md4.1 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Stitch MCP — Edit Screens

Edits one or more existing screens using a text prompt. This is the primary iteration tool — instead of regenerating a screen from scratch (60–180s), you can refine what you already have with targeted changes.

Critical prerequisite

Only use this skill when the user explicitly mentions "Stitch".

You must have both a projectId AND at least one screenId before calling this. If you don't:

  • List screens via stitch-mcp-list-screens to find screen IDs
  • Parse from context: projects/123/screens/456 → projectId: 123, screenId: 456

When to use

  • User wants to modify an existing Stitch screen (change colors, layout, content, style)
  • After generation, user says "make it darker", "change the font", "move the nav to the side"
  • The orchestrator's Step 5b offers "Edit this screen"
  • Any iterative refinement on a Stitch design

Call the MCP tool

{
  "name": "edit_screens",
  "arguments": {
    "projectId": "3780309359108792857",
    "selectedScreenIds": ["88805abc123def456"],
    "prompt": "Change the background to dark mode (#09090B). Make the primary color indigo (#818CF8). Increase the font size of the header to 32px bold.",
    "deviceType": "DESKTOP",
    "modelId": "GEMINI_3_1_PRO"
  }
}

Parameter reference

projectId — numeric ID only, no prefix

✅ "3780309359108792857"
❌ "projects/3780309359108792857"

selectedScreenIds — array of numeric screen IDs

✅ ["88805abc123def456"]
❌ ["projects/123/screens/88805abc123def456"]
❌ ["screens/88805abc123def456"]

Select one or more screens to edit. All selected screens receive the same edit instruction.

prompt — the edit instruction

Apply the same quality bar as generation prompts:

  • Be specific: "Change background to #09090B" not "make it darker"
  • Name exact components: "the header navigation" not "the top part"
  • Include values: hex colors, px sizes, specific content text
  • One focused change per call produces better results than many changes at once

deviceType — optional

Same enum as generate_screen_from_text: MOBILE, DESKTOP, TABLET, AGNOSTIC

modelId — optional

ValueUse when
GEMINI_3_1_PRORecommended — complex layouts, high fidelity
GEMINI_3_FLASHFast iteration, wireframes, simple changes
GEMINI_3_PRODeprecated. Still works but will be removed. Use GEMINI_3_1_PRO instead.

Handling output_components

The response may contain output_components with suggestions:

{
  "outputComponents": [
    { "text": "I've updated the background color and adjusted contrast ratios." },
    { "suggestion": "Would you also like me to update the sidebar to match the dark theme?" }
  ]
}

When you see suggestions:

  1. Present them to the user
  2. If the user accepts: call edit_screens again with the suggestion as the new prompt
  3. This creates a natural refinement loop — keep going until the user is satisfied

Timing

Same as generation: 60–180 seconds is normal.

  • Do NOT retry during this window
  • Do NOT assume failure if it takes > 60 seconds
  • Each call creates a new edit — retries mean duplicate edits

After editing

  1. Re-fetch the screen: stitch-mcp-get-screen with the same projectId and screenId
  2. Show the updated screenshot to the user
  3. Offer:
    • "Continue editing?" → another edit_screens call
    • "Generate variants of this version?" → stitch-mcp-generate-variants
    • "Convert to code?" → framework conversion workflow

Anti-patterns

  • Never send vague edit prompts — "make it better" will produce unpredictable results
  • Never use projects/ID format for projectId or screenId — both must be numeric
  • Never batch unrelated edits — "change the color AND completely redo the layout" works poorly

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Audits Stitch-generated components for WCAG 2.1 AA accessibility issues and applies fixes — semantic HTML, ARIA attributes, keyboard navigation, focus management, and screen reader support.

中文简介正在准备中。当前显示原文说明。

gabelul/stitch-kit452026年7月21日 更新

Adds a purposeful animation layer to Stitch-generated components — CSS transitions, Framer Motion (React/Next.js), or Svelte transitions. Always respects prefers-reduced-motion.

中文简介正在准备中。当前显示原文说明。

gabelul/stitch-kit452026年7月21日 更新

Analyzes a Stitch project's screens and synthesizes a natural-language DESIGN.md — visual atmosphere, color palette with hex values, typography rules, and Stitch-ready prompt snippets. Use this before stitch-loop or any multi-page build to establish design consistency.

中文简介正在准备中。当前显示原文说明。

gabelul/stitch-kit452026年7月21日 更新

Extracts a Stitch design and generates production code artifacts — CSS custom properties with dark mode tokens, a Tailwind v4 @theme block, and a semantic design system document. Run this before framework conversion skills.

中文简介正在准备中。当前显示原文说明。

gabelul/stitch-kit452026年7月21日 更新

Converts a Stitch screen, a local HTML file, or a URL into clean, platform-agnostic HTML5 + CSS — semantic markup, CSS custom properties for theming, dark mode via prefers-color-scheme, mobile-first responsive, zero framework dependencies. Works in browsers, WebViews, Capacitor, and Ionic. Only the Stitch route needs an API key.

中文简介正在准备中。当前显示原文说明。

gabelul/stitch-kit452026年7月21日 更新

Conversational design ideation agent that researches trends, explores visual directions, and refines ideas through adaptive questioning — then produces a rich PRD document and auto-generates Stitch screens. Your design buddy that thinks deeply before designing.

中文简介正在准备中。当前显示原文说明。

gabelul/stitch-kit452026年7月21日 更新

查看 gabelul 的全部技能

报告此技能的问题