Menu
Codex 前端類 Use Case 技能。涵蓋響應式 UI 實作、瀏覽器遊戲開發、 Figma 設計轉碼等工作流。觸發條件：當需要使用 Codex 進行前端 UI 開發、視覺驗證、遊戲建構、設計稿轉碼時啟用。
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Codex 數據分析類 Use Case 技能。涵蓋數據集分析、報告生成、 評分迴圈等工作流。觸發條件：當需要使用 Codex 進行數據分析、 探索性分析、模型建構、報告輸出時啟用。
Codex 工程類 Use Case 技能。涵蓋投影片生成、PR 審查、 程式碼庫上手、API 遷移等工作流。觸發條件：當需要使用 Codex 進行程式碼審查、文件生成、程式碼庫導覽、API 升級時啟用。
Codex 評估類 Use Case 技能。涵蓋 eval-driven improvement loop 工作流。觸發條件：當需要使用 Codex 進行持續品質改進、 評分迴圈、LLM-as-judge 評估時啟用。
Codex 整合類 Use Case 技能。涵蓋 ChatGPT 應用開發、Slack 整合 等工作流。觸發條件：當需要使用 Codex 與第三方平台整合、 建構 ChatGPT 應用、從 Slack 觸發任務時啟用。
Codex 行動端 Use Case 技能。涵蓋 iOS/macOS 原生應用開發、 SwiftUI 架構、CLI 優先工作流。觸發條件：當需要使用 Codex 建構 iOS 或 macOS 原生應用、設定 Swift 專案時啟用。
OpenAI Codex Use Cases 完整知識庫。涵蓋 12 個官方 use case， 包含原始與優化 prompt、工作流程、AGENTS.md 配置範例、 技術棧建議。適用於需要參考 Codex 最佳實踐、撰寫 AGENTS.md、 或設計 Codex 自動化工作流的場景。
| name | codex-frontend |
| description | Codex 前端類 Use Case 技能。涵蓋響應式 UI 實作、瀏覽器遊戲開發、 Figma 設計轉碼等工作流。觸發條件：當需要使用 Codex 進行前端 UI 開發、視覺驗證、遊戲建構、設計稿轉碼時啟用。 |
前端類包含 3 個 use case，覆蓋從 UI 實作到瀏覽器遊戲的完整前端開發場景。 核心特色是使用 Playwright 進行自動化視覺驗證。
本分類包含：
| 步驟 | 動作 | 說明 |
|---|---|---|
| 1 | 檢查截圖 | view_image 分析設計 |
| 2 | 提取設計系統 | 色彩、字型、間距 |
| 3 | 盤點元件 | 重用現有元件 |
| 4 | 建構 Layout | Mobile-first 響應式 |
| 5 | 細節實作 | 動畫、互動狀態 |
| 6 | Playwright 驗證 | 多斷點截圖比對 |
Role: You are a senior front-end engineer with pixel-perfect UI skills.
Context: Implementing the UI from {{screenshot_reference}} in the {{framework}} project.
Task:
1. Inspect the screenshot with `view_image`:
- Extract the color palette (primary, secondary, accent, background, text).
- Identify typography (font family, sizes, weights, line heights).
- Map spacing system (padding, margin, gap patterns).
2. Audit existing project components — reuse before creating new ones.
3. Build the layout:
- Start with mobile layout (min-width: 320px).
- Add tablet breakpoint (768px).
- Add desktop breakpoint (1024px+).
- Use CSS Grid/Flexbox — no absolute positioning unless required.
4. Match visual details:
- Exact border-radius, shadows, and opacity values.
- Hover/focus/active states for interactive elements.
- Smooth transitions (200-300ms ease).
5. Validate with Playwright:
- Take screenshots at 375px, 768px, 1440px widths.
- Visual regression test against the reference.
Constraints:
- Prefer existing design tokens/CSS variables over hardcoded values.
- All interactive elements must be keyboard accessible.
- Images must have alt text.
- No inline styles — use class-based styling.
Output Format:
- Implemented component(s) matching the reference.
- `screenshots/` — Playwright captures at 3 breakpoints.
- Brief diff notes if any deviation from reference was necessary.
Self-Verification:
Run Playwright screenshots at all 3 breakpoints.
Visually compare each against the reference.
Check Lighthouse accessibility score > 90.
| 步驟 | 動作 | 說明 |
|---|---|---|
| 1 | 規劃 | 撰寫 PLAN.md |
| 2 | 選引擎 | Phaser 3 或 PixiJS |
| 3 | 搭建專案 | Next.js + 遊戲引擎 |
| 4 | 核心迴圈 | 輸入 → 更新 → 渲染 |
| 5 | 遊戲機制 | 計分、碰撞、進度 |
| 6 | UI 層 | HUD、選單、結算 |
| 7 | 打磨 | 粒子、震動、音效 |
| 8 | 測試 | Playwright 里程碑驗證 |
Role: You are a game developer specializing in browser-based games with Phaser/PixiJS.
Context: Building a {{game_genre}} game called "{{game_name}}".
Task:
1. Planning Phase:
- Draft `PLAN.md` with: game concept, core mechanics, art style, technical architecture, milestone list.
- Choose engine: Phaser 3 (2D with physics) or PixiJS (lightweight rendering).
- Define the game loop: update → render → input handling cycle.
2. Setup Phase:
- Scaffold Next.js project with chosen engine.
- Configure asset pipeline (sprites, audio, fonts).
- Set up hot-reload for rapid iteration.
3. Build Phase (iterate in this order):
a. Core game loop — player input + basic rendering.
b. Game mechanics — scoring, collision, progression.
c. UI layer — HUD, menus, game over screen.
d. Polish — particles, screen shake, sound effects.
e. Use $imagegen to generate sprite assets when needed.
4. Test at each milestone:
- Use Playwright to validate: game loads, input responds, score updates.
- Performance check: 60fps on mid-range hardware.
Constraints:
- Target 60fps minimum for all animations.
- All game state must be serializable (for save/load).
- Use requestAnimationFrame — never setInterval for game loop.
- Assets must be < 5MB total for fast loading.
Output Format:
- Playable game at localhost:3000.
- PLAN.md with architecture decisions.
- README.md with play instructions and controls.
Self-Verification:
Open the game in browser, play through one complete cycle.
Use Playwright to take a gameplay screenshot for verification.
# Browser Game Conventions
## Build & Run
- `npm run dev` — start dev server
- `npm run build` — production build
- `npm run test` — Playwright tests
## Conventions
- Source layout: `src/game/` for engine code, `src/components/` for UI
- Use ECS pattern for game entities
- All assets in `public/assets/`
- Game state must be serializable to JSON
## Testing
- Playwright E2E for each milestone
- Manual playtest after core loop changes
| 步驟 | 動作 | 說明 |
|---|---|---|
| 1 | 取得設計數據 | get_design_context |
| 2 | 映射 metadata | get_metadata（如需要） |
| 3 | 截圖參考 | get_screenshot |
| 4 | 分析設計系統 | Token 映射 |
| 5 | 實作 UI | Mobile-first |
| 6 | 視覺驗證 | Playwright 多斷點比對 |
Role: You are a front-end engineer converting Figma designs to production code.
Context: Implementing Figma frame {{figma_node_id}} from file {{figma_file_key}} in this {{framework}} project.
Task:
1. Fetch Design Data:
- Call `get_design_context` for the target node/frame.
- If response is truncated, use `get_metadata` then re-fetch specific nodes.
- Run `get_screenshot` for the exact variant before coding.
2. Analyze Design System:
- Extract: colors, typography, spacing tokens.
- Map to existing project design tokens / CSS variables.
- Identify reusable components already in the codebase.
3. Implement:
- Build mobile-first (320px → 768px → 1024px+).
- Reuse existing design system components — do NOT create a parallel system.
- Translate Figma auto-layout to CSS Grid/Flexbox.
- Use Figma's localhost image/SVG sources directly — no placeholders.
4. Validate:
- Side-by-side comparison: Playwright screenshot vs Figma reference.
- Check at 3 breakpoints.
- Iterate until visual match is satisfactory.
Constraints:
- Never invent new icon packages — use Figma-provided SVGs.
- All interactive elements must have hover/focus/active states.
- Keyboard accessibility required.
- Respect existing code conventions.
Output:
- Implemented component(s) matching the Figma design.
- `validation/` — Playwright screenshots at 3 breakpoints.
- `design_mapping.md` — How Figma tokens mapped to project tokens.
Codex 會自動產出不同斷點截圖，與 Figma get_screenshot 的結果並排比較。
| 技術 | 用途 | 相關 Use Case |
|---|---|---|
| Playwright | 視覺驗證 | 全部 |
| React | UI 框架 | frontend-designs |
| Next.js | 遊戲專案框架 | browser-games |
| Phaser 3 | 遊戲引擎 | browser-games |
| PixiJS | 渲染引擎 | browser-games |
| Figma MCP | 設計轉碼 | figma-designs-to-code |
| CSS Grid/Flexbox | Layout | frontend-designs, figma-designs-to-code |