code-ai-installer 1.3.1 → 1.5.0

package/locales/en/.agents/skills/google_stitch_skill/SKILL.md

@@ -0,0 +1,533 @@
+---
+name: google_stitch_skill
+description: Generating UI screens via Google Stitch (AI) with direct access via StitchMCP. Rapid ideation after Approved UX Proposal, prompt engineering, quality control. Tied to the UX/UI Designer gate.
+last_verified: 2026-03-11
+version: 3.0
+mcp_server: StitchMCP
+---
+
+# Skill: Google Stitch — AI UI Generation (MCP-Integrated)
+
+## Purpose
+
+Describes **how and when** the UX/UI Designer uses Google Stitch:
+- for fast generation of initial screen variants via prompts,
+- as a rapid ideation tool between UX Proposal and handoff to Figma,
+- for generating frontend code as a starting point for DEV.
+
+> [!IMPORTANT]
+> Since version 3.0, Stitch is available **directly via MCP** — the agent generates, iterates, and
+> manages screens without manually copying to the browser.
+> Stitch is a **rapid prototyping** tool, not for final design.
+> Any output requires refinement in Figma and a review via `$design_parity_review`.
+
+---
+
+## What is Google Stitch
+
+Google Stitch is an AI tool from Google Labs (launched at Google I/O 2025, based on the acquired Galileo AI). It turns text prompts into UI mockups with export to Figma and HTML/CSS/React code.
+
+**Engine:** Gemini 3 Pro / Gemini 3 Flash.
+**URL:** https://stitch.withgoogle.com
+**Access:** Google account + StitchMCP server connected to the agent.
+
+---
+
+## StitchMCP — API Reference
+
+### Available tools
+
+| Tool | Purpose | Key parameters |
+|------|-----------|-------------------|
+| `create_project` | Create a new project | `title` |
+| `list_projects` | List of all projects | `filter` (owned/shared) |
+| `get_project` | Project details | `name` (projects/{id}) |
+| `list_screens` | List of screens in a project | `projectId` |
+| `get_screen` | Screen details (code, metadata) | `projectId`, `screenId` |
+| `generate_screen_from_text` | 🎯 **Screen generation by prompt** | `projectId`, `prompt`, `deviceType`, `modelId` |
+| `edit_screens` | Editing existing screens | `projectId`, `selectedScreenIds[]`, `prompt` |
+| `generate_variants` | Generating screen variants | `projectId`, `selectedScreenIds[]`, `prompt`, `variantOptions` |
+
+### Models (modelId)
+
+| ID | Description | When to use |
+|----|----------|-------------------|
+| `GEMINI_3_FLASH` | Fast, higher limit | Initial iterations, exploring variants |
+| `GEMINI_3_PRO` | High quality | Final variants, complex layouts |
+
+### Device types (deviceType)
+
+| ID | Description |
+|----|----------|
+| `DESKTOP` | Desktop UIs (dashboards, admin panels) |
+| `MOBILE` | Mobile UIs (iOS/Android) |
+| `TABLET` | Tablet layouts |
+| `AGNOSTIC` | Universal (popups, widgets, modals) |
+
+> [!TIP]
+> For popup/modal/widget templates use `AGNOSTIC` — they are not tied to a specific device.
+> For dashboard pages use `DESKTOP`.
+
+---
+
+## Work modes (via MCP)
+
+| Mode | Model | When |
+|-------|--------|-------|
+| **Fast exploration** | `GEMINI_3_FLASH` | Initial variants, layout exploration, fast iterations |
+| **High-quality generation** | `GEMINI_3_PRO` | Final variant, complex layouts |
+
+> [!WARNING]
+> `generate_screen_from_text` and `edit_screens` can take **up to a few minutes**.
+> **DO NOT repeat the call** — wait for the response. Even on a connection error, generation may finish successfully.
+> Check result: `get_screen` or `list_screens`.
+
+---
+
+## Stitch's place in the UX/UI workflow
+
+```
+PRD → UX Clarification → UX Proposal (Approved)
+                                   ↓
+                    [STITCH MCP: Rapid Ideation]
+                    create_project → generate_screen_from_text
+                    → edit_screens / generate_variants
+                                   ↓
+                     get_screen → Copy to Figma → Polish & Brand
+                                   ↓
+                        UX Spec + Screen Inventory
+                                   ↓
+                        Parity Review ($design_parity_review)
+                                   ↓
+                          Handoff → ARCH / DEV
+```
+
+### When to use Stitch ✅
+- After receiving **Approved** on UX Proposal — for fast materialization of flows into screens
+- To demonstrate to the client/PM "how it might look" before polishing in Figma
+- For generating variants of a single screen (layout exploration)
+- As a foundation for frontend code during fast prototyping
+
+### When NOT to use Stitch ❌
+- For final UI requiring strict adherence to a design system
+- For complex multi-screen flows (Stitch struggles with consistency >2–3 screens)
+- When a ready Figma with components already exists — it's easier to work directly
+- For animations and complex interactive states
+- When the design system is fully defined and pixel-perfect is needed — Figma directly
+
+---
+
+## Inputs
+
+- Approved UX Proposal (mandatory — Stitch is forbidden without it)
+- PRD / User Stories (feature context)
+- Design tokens (colors, fonts, spacing) — to include in the prompt
+- Previous Stitch screens (screenId — for `edit_screens` / `generate_variants`)
+
+---
+
+## Workflow algorithm (MCP)
+
+### Step 1: Project preparation
+
+```
+IF the project already exists:
+  → list_projects → find projectId
+  → list_screens(projectId) → check existing screens
+
+IF no new project exists:
+  → create_project(title: "Project name")
+  → remember projectId from response
+```
+
+### Step 2: Screen generation
+
+```
+generate_screen_from_text(
+  projectId: "<id>",
+  prompt: "<prompt using the checklist below>",
+  deviceType: "DESKTOP" | "MOBILE" | "TABLET" | "AGNOSTIC",
+  modelId: "GEMINI_3_FLASH"   // Flash for the first iteration
+)
+→ remember screenId from response
+→ if output_components contains suggestions — show to the user
+```
+
+### Step 3: Iteration (edit_screens)
+
+```
+edit_screens(
+  projectId: "<id>",
+  selectedScreenIds: ["<screenId>"],
+  prompt: "Move the CTA button to bottom, change color to #4361EE",
+  modelId: "GEMINI_3_FLASH"
+)
+```
+
+### Step 4: Variants (generate_variants)
+
+```
+generate_variants(
+  projectId: "<id>",
+  selectedScreenIds: ["<screenId>"],
+  prompt: "Generate layout variations",
+  variantOptions: { numberOfVariants: 3 }
+)
+```
+
+### Step 5: Getting the result
+
+```
+get_screen(
+  projectId: "<id>",
+  screenId: "<screenId>",
+  name: "projects/<id>/screens/<screenId>"
+)
+→ receive code / metadata of the screen
+```
+
+---
+
+## Prompt engineering for Stitch
+
+### Structure of an effective prompt
+
+```
+[Screen type] + [Platform] + [Content/elements] + [Style] + [Constraints]
+```
+
+**Example of a bad prompt:**
+> "Make a registration form"
+
+**Example of a good prompt:**
+> "Mobile registration screen for a fintech app. Include: email field, password field with show/hide toggle, 'Continue' primary button, 'Sign in' text link below. Dark background (#0F0F14), accent color teal (#00C9A7), Inter font, 16px body. Error state for invalid email shown inline below the field. No illustrations."
+
+### Prompt elements (checklist)
+
+- [ ] **Screen type:** Dashboard / Form / List / Detail / Onboarding / Modal / Popup / Widget
+- [ ] **Platform:** Mobile (iOS/Android) / Web desktop / Tablet / Responsive / Wix iframe embed
+- [ ] **Content:** List specific elements (fields, buttons, cards, charts)
+- [ ] **State:** Default / Loading / Empty / Error / Success / Hover / Disabled
+- [ ] **Style:** color scheme (hex), font, border radius, density, shadows
+- [ ] **Constraints:** what is NOT needed ("No illustrations", "No sidebar", "No gradients on buttons")
+- [ ] **Sizes (if embed):** exact viewport sizes (e.g. "400×500px container")
+
+### Prompts for specific states
+
+**Empty state:**
+> "...empty state screen when user has no data yet. Show a simple icon (no illustrations), headline 'No items yet', subtext 'Create your first item to get started', and a primary CTA button."
+
+**Error state:**
+> "...error/failure screen. Network error message in a toast/snackbar at the bottom, red (#FF4D4F), with retry action. Main content stays visible but dimmed."
+
+**Loading state:**
+> "...loading state with skeleton screens (gray animated placeholders) instead of content. No spinners."
+
+---
+
+## Iterations via edit_screens
+
+After the first generation, use `edit_screens` for refinements:
+
+### Targeted fixes (preferred)
+```
+prompt: "Move the CTA button to the bottom of the screen, fixed position."
+prompt: "Change the card layout to 2 columns instead of 1."
+prompt: "Replace the pie chart with a bar chart."
+prompt: "Change background color to #0F0F14 and accent to #00C9A7."
+```
+
+### Iteration template
+```
+Alignment fix: "[Element] should be [position/alignment]"
+Color fix:     "Change [element] color to [hex]"
+Layout fix:    "Switch [layout A] to [layout B]"
+Content fix:   "Replace [old text/element] with [new text/element]"
+Size fix:      "Make [element] [larger/smaller], [exact px if known]"
+Remove:        "Remove [element] entirely"
+Add:           "Add [element] at [position]"
+```
+
+### Forbidden in iterations
+- Changing the concept entirely (better a new `generate_screen_from_text`)
+- Prompting "make it beautiful" without details — Stitch won't guess
+- More than 5 edits in one `edit_screens` — split into steps
+
+---
+
+## Export to Figma
+
+### When to export
+After 2–3 iterations via `edit_screens`, when the layout is satisfactory.
+
+### Process (via MCP)
+1. `get_screen` → get screen data
+2. Open Stitch UI in the browser → **"Copy to Figma"**
+3. Figma → `Cmd/Ctrl + V` → paste
+
+> [!NOTE]
+> Copy to Figma is currently only available via UI. MCP provides code and metadata,
+> but pasting into Figma requires a manual step via the browser.
+
+### After pasting into Figma — mandatory steps
+- [ ] Replace fonts with project fonts (Stitch uses system ones)
+- [ ] Bind colors to design system tokens
+- [ ] Replace generated components with components from the DS
+- [ ] Check spacing — snap to 4/8px grid
+- [ ] Add missing states (loading/empty/error if not generated)
+- [ ] Check a11y: contrast, touch target sizes
+- [ ] Add a markup in layers: `[Stitch-generated]`
+
+> [!WARNING]
+> Stitch **DOES NOT** create Figma components and DS tokens — everything must be finalized manually.
+
+---
+
+## Code Export (via MCP)
+
+`get_screen` returns HTML/CSS/React code. Use as:
+- Structure reference for DEV
+- Starting point for a fast prototype
+- Checking your own understanding of the layout
+
+### Code retrieval algorithm
+```
+get_screen(projectId, screenId, name)
+→ response contains generated screen code
+→ save to a file with the mark [Stitch-generated scaffold]
+→ pass to DEV
+```
+
+### What NOT to do
+- Pass Stitch-code as production-ready without DEV review
+- Expect components to match shadcn/WDS/other DS
+- Copy inline styles to production — replace with CSS tokens
+
+---
+
+## Stitch Constraints (know and account for)
+
+| Constraint | Impact | Mitigation |
+|-------------|---------|------------|
+| Low consistency between screens | Different styles on different screens | Prompt the identical style explicitly, use `generate_variants` |
+| No animation and transitions | Motion cannot be conveyed | Document separately in UX Spec |
+| No DS-token support | No auto-apply brand | Manual replacement after retrieving code |
+| Complex flows >3 screens — poorly | Loss of consistency | Generate 1–2 screens per project |
+| MCP generation is slow | Up to several minutes | Do not repeat the call, wait for response |
+| Gen limits apply | Same account limits | Plan generation budget |
+| Code is not production-ready | inline-styles, no DS, no a11y | Only as a scaffold, mandatory DEV-review |
+| Copy to Figma — UI only | MCP does not support Figma export | Manual step via browser |
+
+---
+
+## Working with existing projects
+
+### Projects in Stitch
+
+Use `list_projects` to find existing projects and `list_screens(projectId)` for screens.
+
+> [!TIP]
+> If a screen already exists — use `edit_screens` and `generate_variants` instead of creating a new one.
+> Example: `edit_screens(projectId: "<id>", selectedScreenIds: ["<screenId>"], prompt: "...")`
+
+---
+
+## Workflow cards (MCP scenarios)
+
+### Scenario 1: Rapid Screen Ideation
+```
+Goal: Quickly show 2–3 screen variants for approval
+Model: GEMINI_3_FLASH
+Steps:
+1. create_project(title) or select existing
+2. generate_screen_from_text(prompt, GEMINI_3_FLASH) → screenId
+3. generate_variants(screenId, prompt, { numberOfVariants: 3 })
+4. Show variants to the user (get_screen for each)
+5. Select → edit_screens for refinements
+DoD: Layout chosen and approved
+```
+
+### Scenario 2: Iterating an existing screen
+```
+Goal: Improve an existing screen based on feedback
+Model: GEMINI_3_FLASH → GEMINI_3_PRO for final
+Steps:
+1. list_screens(projectId) → find screenId
+2. get_screen → see current state
+3. edit_screens(screenId, prompt with edits, GEMINI_3_FLASH)
+4. Repeat 2–3 times
+5. Final iteration with GEMINI_3_PRO
+DoD: Screen passed 2+ iterations, user approved
+```
+
+### Scenario 3: State Coverage
+```
+Goal: Cover all states of the screen (loading/empty/error/success)
+Model: GEMINI_3_FLASH
+Steps:
+1. generate_screen_from_text → default state
+2. generate_screen_from_text → "Same screen but loading state with skeletons"
+3. generate_screen_from_text → "Same screen but empty state"
+4. generate_screen_from_text → "Same screen but error state with inline message"
+5. All 4 screenId → document in Screen Inventory
+DoD: 4 states generated
+```
+
+### Scenario 4: Component Template
+```
+Goal: Generate/update UI component by design specification
+Model: GEMINI_3_FLASH → GEMINI_3_PRO
+Device: AGNOSTIC (for modal/popup/widget) or DESKTOP (for dashboard)
+Steps:
+1. Take Template Brief / Design Spec
+2. Extract: unique elements, colors, style
+3. generate_screen_from_text(prompt with component + visual elements, deviceType)
+   OR edit_screens(existing screenId) if updating
+4. edit_screens → 2–3 iterations on edits
+5. get_screen → get code for DEV
+6. Verification against $design_parity_review
+DoD: Component visually matches specification, all elements visible
+```
+
+---
+
+## Prompt examples
+
+> Full collection of ready-made prompts: `examples/scr_prompts.md`
+
+### Glassmorphism Modal
+```
+Modal dialog, 400×500px container, centered overlay on a page.
+Glassmorphism style: frosted glass background (rgba(255,255,255,0.15)),
+backdrop-filter: blur(20px), border: 1px solid rgba(255,255,255,0.2).
+Elements: close button (top-right, ×), header image area (top),
+headline "Your Title" (24px bold), body text (14px),
+content area with action button,
+CTA button (full width, gradient accent, rounded 12px).
+Font: Inter. No external illustrations. Dark overlay behind modal.
+Default state.
+```
+
+### Dashboard — Admin Settings Editor
+```
+Web admin dashboard, desktop. SaaS-style panel.
+Left sidebar: navigation links (Dashboard, Content, Settings, Analytics).
+Main content area: Settings editor with live preview.
+Top: section selector (tabs), page title.
+Center: split layout — left: form controls (color pickers, text inputs,
+dropdowns for font/size/position, toggle switches),
+right: live preview of content in a centered frame.
+Bottom: "Save" primary button, "Reset" secondary button.
+Color: white background (#FAFAFA), sidebar dark (#1A1A2E), accent (#4361EE).
+Font: Inter 14px. Border-radius: 8px. Subtle shadow on cards.
+Loaded state with sample data filled in.
+```
+
+---
+
+## Anti-patterns when working with Stitch MCP
+
+- ❌ Repeating `generate_screen_from_text` if the first call "hung" — wait
+- ❌ Using `GEMINI_3_PRO` for initial experiments — start with `GEMINI_3_FLASH`
+- ❌ Handing Stitch code to DEV without the "scaffold only" mark
+- ❌ Generating 20+ screens in one project without structuring
+- ❌ Prompting "make it beautiful" without details — result is unpredictable
+- ❌ Accepting the first generation without `edit_screens` — always 2–3 iterations
+- ❌ Using Stitch BEFORE getting Approved on UX Proposal
+- ❌ Copying Stitch inline-styles to production instead of replacing with DS-tokens
+- ❌ Comparing the implementation with Stitch-output — compare with Figma-final
+- ❌ Deleting/re-creating screens instead of `edit_screens` — iteration history is lost
+
+---
+
+## Definition of Done (Stitch MCP output → Figma)
+
+Stitch-material is considered ready for handoff to Figma when:
+
+- [ ] A final variant is chosen (not a draft)
+- [ ] Passed ≥2 iterations via `edit_screens`
+- [ ] All key screen states covered (default + loading/empty/error/success)
+- [ ] Prompt saved (in deliverables for reproducibility)
+- [ ] `get_screen` confirms correct output
+- [ ] Screen ID is documented in Screen Inventory
+
+## Definition of Done (Figma refinement after Stitch)
+
+Figma-material is considered ready for Handoff when:
+
+- [ ] Fonts → project ones (Inter / SF Pro / project specific)
+- [ ] Colors → bound to DS-tokens
+- [ ] Spacing → 4/8px grid
+- [ ] Components → replaced with DS components (if DS exists)
+- [ ] `[Stitch-generated]` mark in Figma layers for tracking
+- [ ] a11y baseline: contrast ≥4.5:1, touch targets ≥44px
+- [ ] Passed verification via `$design_parity_review` → ≤0 P0
+
+---
+
+## Connection with other skills and pipeline
+
+| UX/UI Stage | Stitch MCP Role | Related skill |
+|------------|-----------------|-----------------|
+| Clarification (step 2) | **Do not use** — no Approved yet | — |
+| UX Proposal (step 3) | Optional: `generate_screen_from_text` for fast wireframe variants | — |
+| Screen Inventory | `generate_screen_from_text` for each screen → `get_screen` | `$design_intake` |
+| Handoff → DEV | `get_screen` → code as scaffold + Figma-final | — |
+| Parity Review | Compare implementation with **Figma**, not Stitch-output | `$design_parity_review` |
+
+> [!CAUTION]
+> **Source of truth is determined by the user** at the UX Clarification stage (mandatory question #2).
+> Options: **(a) Stitch** — parity review against Stitch screenshots via MCP,
+> **(b) Figma** — parity review against Figma mockups, **(c) HTML reference** — parity review against repository files.
+> If the source of truth is not defined — consider it a 🔴 P0 BLOCKER, query the user.
+
+---
+
+## Outputs (Deliverables)
+
+### Stitch Output Package
+- **Project ID** and **Screen IDs** (for reproducibility and iterations)
+- **Prompt** (saved)
+- **Iteration list** (which `edit_screens` were executed)
+- **State coverage:** which states are covered (screenId for each)
+
+### DEV Handoff (if code)
+- **HTML/CSS scaffold** from `get_screen` marked "Stitch-generated, needs DS integration"
+- **Layout reference:** sections, ordering, responsive breakpoints
+- **DO NOT pass:** inline styles as final, colors without tokens
+
+---
+
+## Response format
+
+### Summary
+Brief outcome: what was generated, how many variants, which was chosen. Project ID and Screen IDs.
+
+### Deliverables
+- Stitch Output Package (Screen IDs / code / Figma-export)
+- Prompt (for reproducibility)
+
+### Decisions
+Which variant was chosen and why.
+
+### Risks/Blockers
+- Generation limits (remaining budget)
+- Consistency (if >3 screens)
+
+### Open Questions
+Unresolved questions regarding layout/styling.
+
+### Next Actions (IDs: UX-xx)
+- UX-xx: refinement in Figma
+- UX-xx: verification via `$design_parity_review`
+
+---
+
+## Resources
+
+- **Stitch:** https://stitch.withgoogle.com
+- **Google Blog:** https://developers.googleblog.com/en/stitch-a-new-way-to-design-uis/
+- **MCP Server:** StitchMCP (connected to the agent)
+- **Prompt examples:** `examples/scr_prompts.md`
+- **Generation budget:** `resources/budget_planner.md`
+- **Related skills:** `$design_intake`, `$design_parity_review`

package/locales/en/.agents/skills/google_stitch_skill/agents/claude.json

@@ -0,0 +1,19 @@
+{
+  "name": "google_stitch_skill",
+  "display_name": "Google Stitch",
+  "description": "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP. Rapid ideation after an Approved UX Proposal, prompt engineering, quality control. Tied to the UX/UI Designer gate.",
+  "default_prompt": "Use $google_stitch_skill when the task matches the \"Google Stitch\" skill.",
+  "triggers": [
+    "google_stitch_skill",
+    "google stitch skill",
+    "Google Stitch",
+    "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP"
+  ],
+  "capabilities": [
+    "google",
+    "stitch",
+    "skill"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/google_stitch_skill/agents/copilot.json

@@ -0,0 +1,19 @@
+{
+  "name": "google_stitch_skill",
+  "display_name": "Google Stitch",
+  "description": "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP. Rapid ideation after an Approved UX Proposal, prompt engineering, quality control. Tied to the UX/UI Designer gate.",
+  "default_prompt": "Use $google_stitch_skill when the task matches the \"Google Stitch\" skill.",
+  "triggers": [
+    "google_stitch_skill",
+    "google stitch skill",
+    "Google Stitch",
+    "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP"
+  ],
+  "capabilities": [
+    "google",
+    "stitch",
+    "skill"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/google_stitch_skill/agents/gemini.json

@@ -0,0 +1,19 @@
+{
+  "name": "google_stitch_skill",
+  "display_name": "Google Stitch",
+  "description": "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP. Rapid ideation after an Approved UX Proposal, prompt engineering, quality control. Tied to the UX/UI Designer gate.",
+  "default_prompt": "Use $google_stitch_skill when the task matches the \"Google Stitch\" skill.",
+  "triggers": [
+    "google_stitch_skill",
+    "google stitch skill",
+    "Google Stitch",
+    "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP"
+  ],
+  "capabilities": [
+    "google",
+    "stitch",
+    "skill"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/google_stitch_skill/agents/openai.yaml

@@ -0,0 +1,8 @@
+interface:
+  display_name: "Google Stitch"
+  short_description: "Generation of user interface screens through Google Stitch (artificial intelligence) with direc…"
+  default_prompt: "Use $google_stitch_skill when the task matches the \"Google Stitch\" skill."
+dependencies:
+  tools: []
+policy:
+  allow_implicit_invocation: true

package/locales/en/.agents/skills/google_stitch_skill/agents/qwen.json

@@ -0,0 +1,19 @@
+{
+  "name": "google_stitch_skill",
+  "display_name": "Google Stitch",
+  "description": "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP. Rapid ideation after an Approved UX Proposal, prompt engineering, quality control. Tied to the UX/UI Designer gate.",
+  "default_prompt": "Use $google_stitch_skill when the task matches the \"Google Stitch\" skill.",
+  "triggers": [
+    "google_stitch_skill",
+    "google stitch skill",
+    "Google Stitch",
+    "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP"
+  ],
+  "capabilities": [
+    "google",
+    "stitch",
+    "skill"
+  ],
+  "tools": [],
+  "implicit_invocation": true
+}

package/locales/en/.agents/skills/google_stitch_skill/agents/skill.yaml

@@ -0,0 +1,23 @@
+version: 1
+name: "google_stitch_skill"
+display_name: "Google Stitch"
+description: "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP. Rapid ideation after an Approved UX Proposal, prompt engineering, quality control. Tied to the UX/UI Designer gate."
+default_prompt: "Use $google_stitch_skill when the task matches the \"Google Stitch\" skill."
+triggers:
+  - "google_stitch_skill"
+  - "google stitch skill"
+  - "Google Stitch"
+  - "Generation of user interface screens through Google Stitch (artificial intelligence) with direct access through StitchMCP"
+capabilities:
+  - "google"
+  - "stitch"
+  - "skill"
+tools: []
+invocation:
+  explicit: true
+  implicit: true
+localization:
+  default_locale: "en"
+  available_locales:
+    - "ru"
+    - "en"