Token导航 LogoToken导航

Agent Skill · 待分类

performance(性能)

用于辅助界面设计、视觉规范、排版、配色、布局和交互体验优化。它适合让 Agent 根据产品场景整理页面结构、生成 UI 方案、检查视觉一致性或改进组件层级。使用时需要结合现有品牌、设计系统和用户任务,不应只堆装饰元素;涉及真实页面改动时,应通过截图或浏览器预览检查文本溢出、对齐和响应式表现。

performance 是待分类类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为权限未标注;许可证信息还不明确,暂无第三方扫描结果,按安装量在待分类类的 11,209 个 Skill 里排第 9,504。

总安装 240周安装 10GitHub Stars 3下载 80核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:performance(性能)
来源仓库:https://github.com/phazurlabs/ux-ui-mastery
仓库路径:skills/performance
安装命令:
npx skills add https://github.com/phazurlabs/ux-ui-mastery --skill performance
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/phazurlabs/ux-ui-mastery --skill performance

这个 Skill 是做什么的

这个 Skill 的分类还没最终确定,用法以来源仓库的说明为准。

用于辅助界面设计、视觉规范和交互体验优化。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。适合生成 UI 方案、检查视觉一致性或改进组件层级。使用时需结合品牌、设计系统和用户任务,避免堆砌装饰元素。涉及页面改动时应通过截图或浏览器预览检查文本溢出和对齐。

performance 属于待分类类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/performance

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/phazurlabs/ux-ui-mastery --skill performance;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    权限没有明确标注,第一次使用时只给一个无关紧要的任务,观察它到底读了哪些文件和网络。

常见问题

6 条

performance 怎么安装?

在终端执行 npx skills add https://github.com/phazurlabs/ux-ui-mastery --skill performance。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

performance 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

performance 需要什么权限?

权限级别没有明确标注。第一次使用时建议只给无关紧要的任务,观察它实际读写了哪些内容。

performance 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

performance 用的人多吗?

公开数据是累计安装 240 次,近一周安装 10 次,下载 80 次,GitHub 3 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

performance 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理待分类类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 要求有安全扫描结论才肯安装的团队。
  • 需要厂商级技术支持和 SLA 的生产系统。

待分类类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
infsh-cli

infsh CLI · Codex、Claude

4,374,576382权限未标注未明确未收录
ai-video-generation

AI 视频生成 · Codex、Claude

4,313,376382调用外部服务未明确有提醒
lark-openapi-explorer

云雀 OpenAPI 浏览器 · Codex、Claude

2,064,1209,011需要联网未明确有提醒
lark-workflow-standup-report

Lark 工作流程站立报告 · Codex、Claude

2,041,6328,978需要联网未明确通过
use-my-browser

使用我的浏览器 · Codex、Claude

1,848,49140操作浏览器未明确未通过
performance

性能 · 本页

2403权限未标注未明确未收录

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
Performance, States & Patterns
description
Comprehensive patterns for perceived performance, loading states, skeleton screens, optimistic UI, notification systems, empty states, error states, onboarding flows, and offline-first patterns.
triggers

Performance, States & Patterns

Every interface exists somewhere on a spectrum of states. The difference between an application that feels polished and one that feels fragile lies almost entirely in how it handles the moments *between* the ideal. Loading, emptiness, failure, first contact, disconnection — these transitional states collectively occupy more of a user's lived experience than the "happy path" designers typically obsess over. This skill addresses that full spectrum with rigor, providing concrete patterns, code, and architectural guidance for every in-between moment.

The State Spectrum Philosophy

An interface element moves through a predictable lifecycle of states. Designing for all of them is not optional — it is the baseline for professional-grade UX.

EMPTY -> LOADING -> SKELETON -> PARTIAL -> COMPLETE -> ERROR -> OFFLINE
  |         |          |           |          |          |         |
  v         v          v           v          v          v         v
First-use  Spinner   Shimmer    Streaming   Ideal     Recovery  Cached
guidance   or bar    placeholders  data      render    + retry   fallback

Empty is the starting point: the user has arrived but there is no content yet. This could be a brand-new account, a search with no results, or a list that has been cleared. Empty states are opportunities to guide, teach, and motivate.

Loading is the state of uncertainty. The system is working, but the user cannot yet see results. The perceived duration of this state is more important than its actual duration — perceived performance is a design problem, not just an engineering one.

Skeleton is a specific loading sub-state where placeholder shapes mirror the eventual content layout. Skeleton screens reduce perceived wait time by giving the brain structural information to process, making the transition to real content feel faster and less jarring.

Partial is the state where some data has arrived but not all. Streaming server-side rendering, progressive image loading, and paginated lists all produce partial states. The key design challenge is making partial content usable without misleading users about completeness.

Complete is the happy path — all data loaded, all interactions available. This is the state most mockups depict, but it represents only one moment in the lifecycle.

Error is the state where something went wrong. Network failures, server errors, validation problems, permission issues, rate limits — each error type demands a different response. The goal is always: tell the user what happened, what it means for them, and what they can do next.

Offline is the state where network connectivity is lost. Modern applications must handle this gracefully, serving cached content where possible and queuing actions for later synchronization.

Perceived Performance: The Psychological Foundation

Users do not experience milliseconds — they experience *feelings* about speed. Research consistently shows that perceived performance matters more than actual performance for user satisfaction. A 3-second load that feels fast (because of skeleton screens and progressive rendering) outperforms a 2-second load that feels slow (because of a blank screen followed by a sudden content dump).

The three pillars of perceived performance are:

  1. Immediate feedback — acknowledge every user action within 100ms
  2. Visual continuity — maintain spatial relationships during transitions so the brain can track what is happening
  3. Progressive disclosure of content — show something useful as early as possible, then enhance

Skeleton screens, optimistic UI updates, pre-fetching, and progressive loading are all implementations of these principles. Detailed patterns, code examples, and implementation guidance are provided in the reference material.

See: references/perceived-performance-patterns.md

Notification Architecture: Reaching Users at the Right Moment

Notifications are the interface's voice — the mechanism by which an application communicates with users across time and context. A well-designed notification system respects attention as a finite resource. A poorly designed one trains users to ignore everything, or worse, to uninstall.

The notification taxonomy spans six channels, each with different characteristics:

ChannelPersistenceUrgencyContext
ToastEphemeral (3-8s)Low to mediumIn-app, immediate
Banner/AlertUntil dismissedMedium to highIn-app, contextual
BadgeUntil addressedLow (ambient)In-app, persistent
PushUntil openedMedium to highOS-level, external
Notification CenterPermanent listVariesIn-app, on-demand
Email/SMSPermanentVariesExternal, asynchronous

The architectural challenge is building a unified system where a single event can be routed to the appropriate channel based on urgency, user preferences, and context. A new message from a colleague might appear as a toast if the app is open, a push notification if it is not, and an email if the user is offline for hours — all from a single notification event.

Urgency must be calibrated carefully. If everything is marked critical, nothing is. The "crying wolf" anti-pattern — overusing high-urgency styling — is one of the most common notification design failures.

See: references/notification-system-design.md

First-Run Experience: The Onboarding Spectrum

Onboarding is not a single screen or a tooltip tour. It is a spectrum of patterns that help users move from unfamiliarity to competence. The best onboarding serves the user's goal, not the product's feature tour.

The onboarding spectrum, from least to most intrusive:

  1. Progressive disclosure — features reveal themselves as users naturally encounter them. No explicit onboarding UI at all; the interface teaches by being well-designed.
  2. Contextual hints — subtle, one-time tooltips that appear when a user first encounters a specific feature area.
  3. Checklist model — a visible list of setup tasks (often with endowed progress — starting at "1 of 5 complete" rather than "0 of 5") that guides users through initial configuration.
  4. Interactive sandbox — a safe, pre-populated environment where users can experiment without consequences before committing to real data.
  5. Guided tour — a step-by-step walkthrough with coach marks highlighting specific UI elements. Must be skippable, should be fewer than 5 steps, and should focus on the one or two actions that deliver the product's core value fastest.
  6. Video walkthrough — a short (under 90 seconds) video demonstrating the product's core workflow. Always skippable, always optional.

The pattern chosen should match the product's complexity and the user's likely familiarity with the domain. A developer tool can assume higher baseline literacy than a consumer social app. A complex B2B platform may need a checklist; a simple utility may need nothing beyond a well-designed empty state.

Empty states and onboarding are deeply connected. The first-use empty state is often the user's first onboarding moment — it should answer "What is this?" and "What do I do first?"

See: references/empty-error-onboarding-states.md

Cross-References to Related Skills

This skill intersects heavily with other areas of the UX/UI knowledge base:

  • Component Patterns & Code (component-patterns-code) — Foundational component architecture for buttons, forms, modals, and other elements referenced throughout the state patterns here. Loading buttons, form validation states, and modal error displays all build on component-level patterns.
  • Interaction & Motion Design (interaction-motion-design) — Animation timing, easing curves, and transition choreography for skeleton shimmer effects, toast enter/exit animations, progress indicator motion, and onboarding tour transitions. Motion is the connective tissue that makes state changes feel natural rather than jarring.
  • Mobile UX Design (mobile-ux-design) — Mobile-specific considerations for notification positioning, pull-to-refresh loading patterns, offline handling on unreliable cellular connections, and touch-optimized error recovery. Mobile constraints (smaller screens, intermittent connectivity, OS-level notification rules) shape many of the patterns here.
  • Accessibility & Inclusive Design (accessibility-inclusive-design) — ARIA live regions for dynamic notification announcements, screen reader behavior during loading states, focus management after error recovery, and ensuring skeleton screens do not confuse assistive technology. Every state pattern must be accessible or it is incomplete.
  • Cognitive Psychology & UX (cognitive-psychology-ux) — The psychological foundations underlying perceived performance (the perception of time), notification urgency calibration (attention and interruption science), empty state motivation (goal-gradient effect, endowed progress), and error message framing (loss aversion, self-efficacy).

Guiding Principles

  1. Every state is a design opportunity. Empty states guide. Loading states reassure. Error states teach. Offline states demonstrate reliability. None of these are edge cases — they are core experiences.
  1. Perceived speed trumps actual speed. Invest in skeleton screens, optimistic updates, and progressive loading. Users remember how fast something *felt*, not how fast it was.
  1. Notifications are a finite resource. Every notification you send withdraws from a trust account. Send too many, and users revoke permission or uninstall. Earn the right to interrupt.
  1. Error messages are a conversation. Write them in plain language. Tell the user what happened, why it matters to them, and what they can do about it. Never show raw error codes or stack traces.
  1. Onboarding should be invisible when possible. The best onboarding is an interface so clear it does not need a tour. When explicit onboarding is necessary, make it serve the user's first goal, not the product's feature list.
  1. Offline is not an error — it is a state. Modern applications should handle disconnection as a normal operating condition, not an exception. Cache aggressively, queue writes, and synchronize gracefully.

State Design Audit Checklist

When designing any new feature, view, or component, walk through this checklist to ensure every state has been considered. Incomplete state coverage is the single most common source of "unpolished" feelings in production software.

For every data-dependent view, verify you have designed:

  • First-use empty state — What does a new user see before creating any content? Is there an illustration, explanation, and CTA? Does the empty state educate or at least orient?
  • No-results state — If the user searches or filters and nothing matches, what do they see? Are there suggestions for broadening the search?
  • Loading state — What does the user see during the data fetch? Is it a skeleton (preferred for content-heavy views) or a spinner (acceptable for brief waits)? Is the loading indicator accessible to screen readers?
  • Partial loaded state — If data arrives in stages (paginated, streamed), does the partial view look intentional? Can users interact with the content that has arrived?
  • Complete state — The standard "happy path" design. This is what most mockups show.
  • Error state — What happens if the data fetch fails? Is there a retry button? Is the error message human-readable? If cached data exists, is it shown with a staleness indicator?
  • Offline state — If the user loses connectivity, what degrades? What remains functional? Is there a clear indicator that the user is offline?
  • Permission-required state — If the view depends on a user permission (location, notifications, camera), what does the user see before granting it?

For every user action, verify you have designed:

  • Immediate feedback — Does something visible change within 100ms of the action? This could be a button state change, a spinner, or an optimistic update.
  • Success confirmation — How does the user know the action succeeded? Is it a toast, an inline change, a redirect?
  • Failure handling — If the action fails, what does the user see? Is there a retry path? Is their input preserved?
  • Undo mechanism — For destructive or significant actions, is there a brief undo window?

For every notification event, verify you have designed:

  • Channel routing — Which channel(s) will deliver this notification? Does it change based on user context (in-app vs. away)?
  • Urgency classification — Is the urgency level appropriate? Would a user agree that this level of interruption is warranted?
  • Preference respect — Can the user control whether they receive this notification, and through which channel?
  • Accessibility — Does the notification announce itself to screen readers via the appropriate ARIA live region?

This checklist is not exhaustive, but completing it for every new feature will eliminate the vast majority of state-related UX gaps that erode user trust over time.

平台分布

Codex

33.53%
按下载量换算27

Claude

28.51%
按下载量换算23

Cursor

20.07%
按下载量换算16

Gemini CLI

8.58%
按下载量换算7

安全审计

暂无安全审计结果可展示。

权限和风险

权限未标注

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills