owwzo/ai-agent/openui-interactive-reliability

Claude Code and agent skill by owwzo/ai-agent

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Build and review reliable openui-lang interactive interfaces. Use whenever a user asks for a quiz, planner, calculator, map, dashboard, form, checklist, or any UI with clickable controls, editable fields, or state changes. Prevent disabled controls, dead inputs, invalid positional arguments, and state updates that do not visibly change the interface.

Install

npx vskill@latest install owwzo/ai-agent/openui-interactive-reliability

SKILL.md

Reliable OpenUI Interfaces

Design the interface around a small, explicit state model and verify every interactive control before presenting it.

Core rules

  1. Put root = ... first for progressive rendering.
  2. Every $variable must be declared with a sensible default and must affect a rendered component or action.
  3. Every input must either change visible UI state or be inside a Form whose submit button sends a concise @ToAssistant(...) request carrying the form values.
  4. For Button, respect positional arguments. The optional disabled argument comes near the end; do not append true casually. If the button should work, omit disabled or pass false explicitly.
  5. Do not use a dead Input or Select: bind it to a $variable that is read by displayed text, computed data, or submit context.
  6. Use Action([...]) for multiple updates and ensure each @Set changes declared state. After a click, the visible question, score, selected marker, itinerary, or result should change.
  7. For dynamic lists, use @Each with stable item fields and ensure generated buttons are not disabled unless there is a real condition.
  8. For maps and images, use the extended components only after loading the UI component catalog. Use MapMarker coordinates appropriate to the displayed city and SearchImage with precise 2–4 word queries.

Preflight checklist

Before outputting an openui block, inspect it as if clicking every control:

  • Can each button receive a click? Check the disabled positional argument.
  • Does each button action execute a state change or a valid @ToAssistant?
  • Is every form submit button enabled and connected to the form?
  • Are all input bindings declared and used?
  • Does a segmented control/select visibly update the content it controls?
  • Does a map marker selection visibly update selected text or details?
  • Are all referenced names defined and all defined names reachable from root?
  • Are component arguments positional and in the documented order?

Common failure patterns

  • A final true intended as block is accidentally interpreted as disabled. Prefer named-by-position discipline: supply only arguments needed, or use false for disabled and true only for block after confirming the signature.
  • Quiz answers increment the question but do not update the score because the answer comparison is wrong. Compare the clicked option with the stored answer before @Set.
  • A submit button is disabled by default “for safety,” leaving no path to completion. Make it enabled unless validation is actually implemented.
  • A form collects values but the assistant prompt ignores them. Refer to the form context implicitly and ask for the resulting plan; do not duplicate values in the message.

HtmlView 游戏界面

对于可玩的小游戏,优先使用单个 HtmlView 包含 canvas、样式和内联 JavaScript;保持游戏状态集中在少量变量中,并让每次输入都产生可见变化(位置、分数、生命、提示或模式)。

  • 同时提供触控和键盘输入;触控按钮使用 pointerdown/pointerup,并在 pointercancel、pointerleave 与全局 pointerup 时释放按键,避免“卡住”。
  • 对按住有效的动作(移动、蓄力、连发)使用按键状态;对开始、重置、开火等单次动作使用独立函数,避免同一事件既触发绑定处理又触发额外监听而重复执行。
  • 在 requestAnimationFrame 循环中限制 dt,并在 mode !== "play" 时停止游戏逻辑;绘制层仍应显示标题、暂停、失败和胜利遮罩。
  • 每个可交互控件都要能在移动端点击,且提供明确的重试/重置路径;不要把按钮标记为 disabled,除非确有可见且可恢复的条件。
  • 在交付前静态预检:root 是首行;HTML 字符串中的引号已正确转义;canvas 尺寸与 CSS 比例一致;开始、暂停/重置、移动、跳跃/射击、结束状态均有可达路径;不重复注册同一按钮的同类事件。

Output style

Use Chinese when the user communicates in Chinese. Keep the surrounding explanation brief. For an interactive request, return one valid openui-lang block and no extra pseudo-code. If the interface depends on unknown component signatures, call get_ui_components first.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/owwzo/ai-agent/openui-interactive-reliability/badge)](https://verified-skill.com/skills/owwzo/ai-agent/openui-interactive-reliability)

Something wrong with this skill or its page? Tell us.