Skip to content
cccskills
FreePublished on GitHub

ui-wireframe-to-html

PRD 到 UI 线框 / 结构阶段:当需要把 PRD 的界面定义先转成 screen inventory、状态模型、ASCII 布局、低保真 HTML mockup, 并先确认结构/状态而不是视觉 polish 时使用。

这个 Skill 只负责 UI mockup 的结构阶段;如果用户要高保真、真实前端项目对齐、 project-native preview、component map、implementation notes 或“开发完全复刻”,必须转用 ui-mockup-desktop-workbench。

The English summary is being prepared. Showing the original description.

Run score Not tested

See how to install

Files included (7)

  • SKILL.md8.0 KB
  • evals/evals.json5.4 KB
  • references/templates/ascii-layout.md931 B
  • references/templates/screen-inventory.md421 B
  • references/templates/state-model.md663 B
  • references/templates/wireframe-handoff.md553 B
  • scripts/check_wireframe_package.py859 B

SKILL.md (original)

Review the instructions the agent will receive before you install.

ui-wireframe-to-html

中文速查

  • 中文名:PRD 到 UI 线框 / 结构阶段
  • 英文稳定名:ui-wireframe-to-html
  • 分类:产品与 PRD
  • 你可以这样叫我:把 PRD 变成线框、先出 UI 结构、先做 ASCII 布局、只确认页面结构和状态、生成低保真 HTML wireframe
  • 适合:从 PRD 界面定义输出 screen inventory、状态模型、ASCII layout 和低保真 HTML mockup,先明确结构和状态再决定是否进入高保真。
  • 不适合:直接做高保真视觉;让前端复刻 UI;对齐真实项目组件、token、icon;输出 component map 或 implementation notes。这些场景用 ui-mockup-desktop-workbench。

Relationship To ui-mockup-desktop-workbench

This Skill is the structure-only stage of the broader UI mockup workflow.

Use this Skill when the user explicitly wants to stop at structure, wireframe, information architecture, state model, or low-fidelity HTML.

Use ui-mockup-desktop-workbench instead when the user wants:

  • high-fidelity visual handoff
  • real frontend project alignment
  • project-native preview route/component
  • production component mapping
  • implementation notes
  • screenshots for final UI approval
  • "开发完全复刻这个 UI"

The normal staged path is:

PRD -> ui-wireframe-to-html -> structure confirmation -> ui-mockup-desktop-workbench -> visual handoff / preview -> implementation

When ui-mockup-desktop-workbench is already active, do not call this Skill separately. The main Skill includes this structure stage internally.

Overview

将 PRD 中的界面定义逐步推进为:

  1. Screen inventory
  2. State model
  3. ASCII layout
  4. Optional low-fidelity HTML wireframe

The purpose is to make the page structure, state coverage, and interaction skeleton reviewable before anyone spends effort on visual polish.

Context Intake

优先确认或读取:

  • 关联 PRD、页面需求、草稿或 UI 目标。
  • 当前页面在全链路中的位置。
  • 目标屏幕范围:主页面、列表、详情、弹窗、设置、异常状态等。
  • 已有结构定义、流程图、状态说明或截图。
  • 是否存在 docs/templates-local/ 下的 UI override。
  • 用户是否明确要求只做结构阶段,还是最终要进入高保真。

If the user provides enough PRD context, proceed with assumptions instead of asking broad questions. Ask only when the missing answer changes page scope or state coverage.

Responsibilities

这个 Skill 负责:

  1. 读取关联 PRD 和当前模块定位。
  2. 明确页面在全链路中的位置。
  3. 提取 screen inventory。
  4. 补齐关键状态模型。
  5. 输出 ASCII 结构图。
  6. 需要时生成低保真 HTML mockup。
  7. 列出进入高保真前必须确认的问题。

它不负责:

  • 跳过结构层直接出视觉稿。
  • 脱离 PRD 自行发明新交互。
  • 用 HTML mockup 替代 UI 结构说明。
  • 声称低保真 HTML 可直接复用为生产前端。
  • 输出高保真视觉、project-native preview、component map 或 implementation notes。

Workflow

  1. Read PRD and scope

    • Identify the product goal, target users, workflow stage, and page boundaries.
    • If the PRD is missing core goals or workflow, route to prd-architect.
    • If the PRD has conflicts or cannot support a screen model, route to prd-review.
  2. Build screen inventory

    • List pages, panels, modals, popovers, drawers, and persistent shell areas.
    • Mark each item as required, optional, or assumption.
    • Trace each screen to PRD text, UI goal, screenshot, or explicit assumption.
  3. Build state model

    • Cover relevant states: empty, draft, loading, queued, running, stopped, success, partial failure, error, permission/login, settings, and recovery.
    • For each state, include trigger, visible regions, primary actions, next state, and recovery path.
    • Do not hide failure states in generic toast-only descriptions.
  4. Draft ASCII layout

    • Draw the primary layout before visual styling.
    • Show fixed rails, flexible regions, scroll containers, artifact/detail panels, and modal overlays.
    • Include layout variants when a state changes the structure.
  5. Optional low-fidelity HTML

    • Create mockup.html only when the user asks for a browser-openable wireframe or when visualizing structure materially helps review.
    • Keep it low fidelity: neutral styling, semantic regions, clear labels, and visible state examples.
    • Label it as structure reference, not production code.
  6. Create structure review notes

    • List confirmed structure decisions.
    • List assumptions.
    • List blockers or questions before high fidelity.
    • Recommend ui-mockup-desktop-workbench when the structure is ready for high-fidelity handoff.

Output Shape

Default outputs:

  • README.md or concise handoff note: source PRD, scope, assumptions, and how to review.
  • screen-inventory.md: screens, regions, source trace, and assumptions.
  • state-model.md: states, triggers, actions, recovery, and PRD trace.
  • ascii-layout.md: desktop or page layout structure.
  • mockup.html: optional low-fidelity HTML wireframe.

If the user asks for a single response instead of files, include the same sections inline.

Rules

  1. Structure before visual polish.
  2. State model before HTML.
  3. PRD traceability before new interaction ideas.
  4. Low-fidelity HTML is optional and never the source of truth.
  5. If PRD has draw.io -> SVG总图或核心流程图,UI 文档必须承接其阶段位置。
  6. If project provides docs/templates-local/ UI overrides, use the local override for structure naming and layout expectations.
  7. If the user asks for "高保真", "真实项目", "组件映射", "前端复刻", or "开发交付", route to ui-mockup-desktop-workbench.

Template References

优先使用项目内 docs/templates-local/ override;没有本地 override 时读取本 Skill 的模板:

  • references/templates/screen-inventory.md
  • references/templates/state-model.md
  • references/templates/ascii-layout.md
  • references/templates/wireframe-handoff.md
  • scripts/check_wireframe_package.py

Definition Of Done

完成标准:

  • UI 结构先于视觉被确认。
  • Screen inventory 覆盖目标页面/区域。
  • 状态模型不缺关键状态和恢复路径。
  • ASCII layout 展示布局区域、滚动行为和主要动作。
  • 输出物至少包含 README/handoff、screen inventory、state model 和 ASCII layout。
  • 需要时才生成低保真 HTML mockup。
  • 明确说明下一步是否进入 ui-mockup-desktop-workbench。

Evaluation

Smoke prompts:

  • 基于这份 PRD 先出 UI 结构、状态模型和 ASCII 布局,不要做高保真。
  • 把这个页面需求转成低保真 HTML wireframe,重点确认状态和布局。
  • PRD 已有页面结构时,先帮我补 screen inventory 和 state model。

Non-trigger prompts:

  • 基于 PRD 出高保真 mockup,开发要完全复刻。
  • 做真实项目 preview,并输出 component map。
  • 帮我写一份 PRD。
  • 只审一下这个 PRD 有没有问题。

Capability checks:

  • The output does not skip directly to visual styling.
  • The output includes screen inventory, state model, and ASCII layout.
  • Optional HTML is labeled as low-fidelity structure reference.
  • The output routes high-fidelity or implementation-oriented requests to ui-mockup-desktop-workbench.

Catalog Notes

  • Category: product-prd
  • Chinese name: PRD 到 UI 线框 / 结构阶段
  • Status: active
  • Relationship: structure-stage companion to ui-mockup-desktop-workbench.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Agent trace 诊断器:当用户提供 trace、日志 JSON、agent 执行记录、工具调用序列,或说“看下这个日志/trace,定位根因”“只说明问题不要改文件”时使用;用中文输出核心根因、可视化执行流程图、trace 证据链、可能文件/行和修复建议,默认只读。不用于没有日志证据的新功能设计、普通代码解释或直接修复代码。

The English summary is being prepared. Showing the original description.

PANGKAIFENG/ai-product-manager-skills16Updated Aug 31, 2026

协作校准 / 认知校准 / 问题脑暴:当用户有模糊感受说不清问题、方案越补越复杂、想挑战假设、 想重新定义问题、想知道成熟领域里怎么叫这个问题时使用。 可用中文唤起:"帮我想想""先聊一下""一起脑暴""我感觉有问题但不清楚是什么" "先别执行,帮我看清问题""挑战我的假设""这个方案是不是想错了""帮我做认知校准" "双向钢人""钢人论证一下""正反最强论证,给我明确判断"。 也适用于:用户直接说"帮我做 X 功能""接入 Y 平台"但背景复杂、或对话早期感觉方向可能跑偏时, AI 主动识别并建议进入脑暴模式。 这里的“挑战假设”主要指问题定义、目标、约束和领域定位层面的假设;如果问题已确认且用户已有具体方案、 计划、架构或决策要压力测试,应转用 grill-me。不用于翻译、摘要、格式整理、单文件小改等简单执行任务。

The English summary is being prepared. Showing the original description.

PANGKAIFENG/ai-product-manager-skills16Updated Aug 31, 2026

AI 工作资产化诊断器 / Assetization Router:当用户提供 AI 协作会话、重复任务、prompt、工作流描述、 团队 AI 使用场景或一次成功交付过程,并要求判断应该沉淀成 Prompt、Context Pack、Workflow、Skill、 Loop、System,或根本不值得沉淀时使用。它只做诊断、分层和下一步资产建议,不替代具体 Skill 创建、 PRD 起草、调研、实现或自动化执行。不用于普通事实查询、一次性文案、trace 根因分析或高责任专业判断。

The English summary is being prepared. Showing the original description.

PANGKAIFENG/ai-product-manager-skills16Updated Aug 31, 2026

设计脑暴 / 实现前方案校准:当用户想把已基本成立的想法、功能方向或产品问题,在写 PRD、画 mockup 或进入开发计划前,先比较方案、确认取舍、对齐 UI/视觉约束,并收敛成可执行设计 spec 时使用。 可用中文唤起:“先脑暴一下方案”“先不要写 PRD,帮我设计几种路径”“参考 brainstorming 把这个需求变成设计 spec” “实现前先讨论设计”。问题还没定义清楚时先用 ai-collaboration-calibration;已有方案要压力测试时用 grill-me; 直接写 PRD 时用 prd-architect;grill-me 返回精确设计 gap 时,本 Skill 只输出 Design Delta。

The English summary is being prepared. Showing the original description.

PANGKAIFENG/ai-product-manager-skills16Updated Aug 31, 2026

竞品决策分析 / Competitive Decision Brief:当用户要调研竞品、替代方案、新产品、SaaS/AI 产品或市场信号, 并希望把官网、定价、评论、文档、更新日志、登录态走查、截图或浏览器取证转成产品定位、路线图、定价、 功能优先级、差异化、Go/No-Go 或 PRD 输入时使用。核心行为是先锚定产品决策,再选择证据渠道, 最后输出可行动的 Product Decision Brief;不要把它降级成泛泛功能清单、纯 UX 走查或“点完所有按钮”报告。

The English summary is being prepared. Showing the original description.

PANGKAIFENG/ai-product-manager-skills16Updated Aug 31, 2026

复杂探索资产化 / Complex Exploration:当用户面对复杂、不确定、多轮迭代的产品策略、Roadmap、商业化定价、竞品定位、复杂 PRD 前置探索、项目复盘或方法论沉淀任务时使用。它先判断任务类型,暴露隐含假设,重构真正问题,规划探索路径和中间产物,并在结束后沉淀认知、结构、方法论、工具和影响力资产。不用于简单润色、翻译、摘要、明确执行任务;问题只需早期认知校准时优先用 ai-collaboration-calibration,已有方案要压力测试时用 grill-me,系统专题研究用 research-topic-compiler。

The English summary is being prepared. Showing the original description.

PANGKAIFENG/ai-product-manager-skills16Updated Aug 31, 2026

See all skills by PANGKAIFENG

Report a problem with this skill