Token导航 LogoToken导航

Agent Skill · 研究检索

design-ops(设计操作)

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。

design-ops 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为操作浏览器;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 52,216。

总安装 563周安装 23GitHub Stars 4下载 182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:design-ops(设计操作)
来源仓库:https://github.com/cuellarfr/design-skills
仓库路径:skills/design-ops
安装命令:
npx skills add https://github.com/cuellarfr/design-skills --skill design-ops
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/cuellarfr/design-skills --skill design-ops

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

提升设计团队效能的运营流程,涵盖冲刺、交接与质量保证。基于 Jake Knapp 的 5 日设计冲刺方法论,适用于高风险决策前验证。使用时需判断问题清晰度与团队共识度是否适合启动冲刺。

安装方式:GitHub,命令为 npx skills add https://github.com/cuellarfr/design-skills --skill design-ops。注意:强调仪式感与文档化,确保知识资产可持续传递。

来源仓库内路径:skills/design-ops

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/cuellarfr/design-skills --skill design-ops;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

    它需要操作浏览器,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

design-ops 怎么安装?

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

design-ops 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。

design-ops 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

design-ops 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。

design-ops 用的人多吗?

公开数据是累计安装 563 次,近一周安装 23 次,下载 182 次,GitHub 4 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。

design-ops 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

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。
  • 第三方扫描已通过,常见的高风险行为没有命中。

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。
  • 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。
  • 需要处理研究检索类任务、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。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
design-ops

设计操作 · 本页

5634操作浏览器未明确通过

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

Design Ops & Handoff

Operational processes that make design teams effective — sprints, handoff, rituals, documentation, and quality assurance.


Design Sprints

A 5-day process for solving big problems and testing ideas before committing engineering resources. Based on Jake Knapp's Sprint methodology developed at Google Ventures.

When to Sprint

SignalSprint Is RightSprint Is Wrong
Problem clarityBig problem, unclear solutionSolution already known — just build it
StakesHigh — wrong direction costs monthsLow — easy to course-correct later
Team alignmentDisagreement on approachTeam already aligned
Time pressureNeed answers this weekCan run a longer discovery process
ComplexityCross-functional problemSingle-discipline task

The 5-Day Structure

DayFocusKey ActivityOutput
MondayMapDefine the challenge, set a 2-year goal, map the problem space, pick a targetSprint questions, problem map, target area
TuesdaySketchReview existing solutions, each person sketches solutions individuallySolution sketches (one per person)
WednesdayDecidePresent sketches, vote (dot voting + decider picks), create storyboardStoryboard for prototype
ThursdayPrototypeBuild a realistic facade — just enough to testTestable prototype
FridayTest5 user interviews, observe together, identify patternsGo/no-go decision with evidence

Sprint Team

RoleCountResponsibility
Decider1Has authority to make final calls. Usually PM or founder
Facilitator1Runs the process, keeps time, manages energy
Designer1-2Leads sketching, builds prototype
Engineer1-2Feasibility gut-checks, helps prototype
Domain expert1-2Customer support, sales, marketing — whoever knows the user
Total5-7More than 7 is counterproductive

Sprint Rules

  1. No devices during working sessions — full attention
  2. Work alone together — individual sketching, not group brainstorming
  3. The Decider decides — democracy is slow; one person breaks ties
  4. Prototype, don't build — facade quality, not production quality
  5. 5 users is enough — you'll see 85% of usability issues with 5 participants

Design-to-Development Handoff

Handoff is the #1 source of quality loss in product teams. The goal: engineers should never have to guess.

Handoff Maturity Levels

LevelDescriptionSymptom
1. ScreenshotDesigner sends a static imageEngineers guess spacing, states, behavior
2. Annotated mockupRedlines, spacing calloutsBetter, but still missing states and edge cases
3. Interactive specFigma/design tool with inspect mode, component linksEngineers can extract values, but behavior is undocumented
4. Full specAll states, transitions, error handling, responsive behavior, tokensEngineers build exactly what was designed
5. IntegratedDesign and code share a component system; handoff is a conversation, not a documentMinimal spec needed — shared language does the work

Target: Level 4 minimum. Level 5 for mature teams.

What a Complete Handoff Includes

Every screen or component handoff must cover:

CategoryRequired Items
LayoutSpacing (in tokens, not pixels), alignment, responsive breakpoints, container behavior
ContentFinal copy (not lorem ipsum), character limits, truncation rules, localization notes
StatesDefault, hover, focus, active, disabled, loading, empty, error, success — for every interactive element
BehaviorWhat happens on click/tap, keyboard interaction, drag, swipe. Transition details
ResponsiveHow the layout adapts at each breakpoint. What reflows, stacks, hides, or changes
TokensColor tokens, typography tokens, spacing tokens — never raw hex/px values
AccessibilityFocus order, ARIA roles, screen reader announcements, touch targets, color contrast
Edge casesLong strings, empty data, error recovery, offline behavior, permissions
AssetsIcons (SVG), images (with alt text), animations (specs or files)

Handoff Anti-Patterns

Anti-PatternProblemFix
"It's in Figma"No context, engineer must reverse-engineer intentWrite behavior notes on every frame
Happy path onlyOnly the ideal flow is designedDesign all states before handoff
Pixel specs"24px left margin"Use tokens: spacing-md
Verbal handoff"I'll explain it in standup"If it's not written down, it doesn't exist
Big-bang handoffEntire feature dropped on engineering at onceHand off in slices that match sprint stories
No error states"We'll figure out errors later"Error states are designed first, not last

Team Rituals and Cadences

Core Design Rituals

RitualFrequencyDurationPurposeWho Attends
Design critiqueWeekly45-60 minImprove work-in-progress through structured feedbackAll designers
Design-eng syncWeekly30 minAlign on upcoming work, resolve handoff questionsDesign lead + eng lead
Design reviewPer milestone30-60 minStakeholder sign-off on directionPM, design, eng leads, stakeholders
Handoff sessionPer story/feature30-45 minWalk engineering through specs, answer questionsDesigner + implementing engineers
Design QAPer release60-90 minVerify implementation matches specDesigner + engineer
RetroBi-weekly or per sprint45 minImprove the processWhole product team
Design sprintAs needed5 daysSolve big problems, validate risky ideasCross-functional team (5-7)

Running Effective Design Critiques

Format: Presenter shares context (2 min) → Group reviews silently (3 min) → Structured feedback (15-20 min per piece) → Next steps (2 min)

Critique rules:

  1. Presenter states what feedback they want — "I need help with the empty state" not "What do you think?"
  2. Feedback references principles, not preferences — "This violates Fitts's Law" not "I don't like the button placement"
  3. Describe problems, don't prescribe solutions — "Users might miss this action" not "Make the button bigger"
  4. No defending — presenter listens, asks clarifying questions only
  5. Written notes — someone captures all feedback items with action owners

Handoff Session Structure

  1. Context (5 min) — What problem does this solve? What user story?
  2. Walkthrough (10-15 min) — Designer walks through the flow, screen by screen
  3. States and edge cases (10 min) — Hover, error, empty, loading, responsive
  4. Tokens and components (5 min) — Which design system components to use, which tokens
  5. Questions (10 min) — Engineers ask anything unclear
  6. Action items — What needs clarification before build starts

Documentation Standards

File Organization

project-name/
├── exploration/          # Early concepts, discarded directions
│   └── v1-concepts/
├── current/              # Active, approved designs
│   ├── flows/            # User flows and journey maps
│   ├── screens/          # Screen-level designs
│   └── components/       # Component-level specs
├── assets/               # Production-ready exports
│   ├── icons/
│   └── images/
├── handoff/              # Handoff documentation
│   └── [feature-name]/
└── archive/              # Past versions (never delete, move here)

Naming Conventions

ElementConventionExample
Files[project]-[feature]-[variant]taskpilot-onboarding-v2
Pages/frames[flow]-[step]-[state]checkout-payment-error
ComponentsMatch code component names exactlyButtonPrimary, not Main CTA
VersionsSemantic: v1.0, v1.1, v2.0Major = new direction, minor = iteration
LayersDescriptive, no "Group 1" or "Frame 47"header/nav/menu-trigger

Version Control Rules

  1. Never overwrite — create a new version, archive the old one
  2. Name the change — "v2.1 — added error states" not just "v2.1"
  3. Link to decisions — every major version links to the decision that drove it (Slack thread, meeting note, sprint review)
  4. Single source of truth — one canonical file per feature. No "copy of" files
  5. Archive, don't delete — move outdated designs to archive with a date stamp

Design QA

Design QA is the process of verifying that what was built matches what was designed. Do it before every release.

Design QA Checklist

CategoryCheckTool
Visual accuracyColors match tokens. Spacing matches spec. Typography is correctSide-by-side comparison, browser dev tools
StatesAll states render correctly: default, hover, focus, active, disabled, error, loading, emptyManual testing, Storybook
ResponsiveLayout works at all breakpoints. Nothing overflows, truncates wrong, or disappearsResize browser, test on real devices
ContentReal content fits. Long strings handled. Empty states designed. Truncation worksTest with edge-case content
InteractionAnimations match spec. Transitions are smooth. Timing feels rightManual testing, slow-motion recording
AccessibilityFocus order correct. Screen reader announces correctly. Contrast passes. Touch targets 44px+axe, keyboard testing, VoiceOver/NVDA
Cross-browserRenders correctly in Chrome, Firefox, Safari, EdgeBrowserStack or manual
PerformanceImages optimized. No layout shift. Animations don't drop framesLighthouse, Core Web Vitals

Filing Design QA Issues

Every issue needs:

**What:** [Element name] on [screen/page]
**Expected:** [What the design spec shows — link to frame]
**Actual:** [What the build shows — screenshot]
**Severity:** [Critical / Major / Minor / Cosmetic]
**Environment:** [Browser, viewport, OS]

Severity scale:

  • Critical — Functionality broken. User can't complete the task
  • Major — Wrong behavior. User can complete the task but with confusion or errors
  • Minor — Visual discrepancy. User isn't affected functionally but quality feels off
  • Cosmetic — Nitpick. Pixel-level differences that only designers notice

Acceptance Criteria for Design

Add these to every user story:

Design acceptance criteria:
- [ ] All states match design spec (link to Figma frame)
- [ ] Responsive behavior verified at 320px, 768px, 1024px, 1440px
- [ ] Keyboard navigation works (Tab, Enter, Escape, Arrow keys)
- [ ] Screen reader announces all interactive elements correctly
- [ ] Design tokens used (no hardcoded colors, spacing, or typography)
- [ ] Loading and error states implemented
- [ ] Edge-case content tested (empty, long strings, special characters)

Process Metrics

Track these to measure design ops health:

MetricWhat It MeasuresTargetHow to Measure
Handoff completeness% of handoffs with all states, tokens, responsive specs>90%Audit handoff docs quarterly
Design QA pass rate% of stories passing design QA on first review>80%Track QA issues per story
Rework rate% of design work that gets redone after handoff<15%Count post-handoff design changes
Sprint velocity impactHow often missing specs block engineering0 blocks/sprintTrack blocker tags in standup
Design debtKnown visual/interaction inconsistencies in productionDecreasing trendMaintain a design debt backlog
Critique participationDesigners bringing work to critique regularly100% monthlyTrack critique presenters

Scaled-Down Operations for Lean Teams

The processes above assume a dedicated design team. But many organizations have a single designer — or someone who does design as part of a broader role. The principles still apply; the execution scales down. This section provides adapted methods for resource-constrained teams, drawing on Leah Buley's UX team-of-one framework.

The Core Mindset

  1. Focus on small wins. You can't transform the organization overnight. Pick one visible improvement, deliver it, and let the result build momentum for the next
  2. Build a user-centered culture, not a UX department. Your goal isn't to own all design work — it's to make everyone on the team think about users
  3. Develop a growth mindset. Seek opportunities to expand skills incrementally. You don't need permission to start practicing research or testing

Guerrilla Methods

When you don't have budget, time, or formal research infrastructure:

Full ProcessGuerrilla AlternativeWhen It's Enough
Formal usability study (lab, recruiting, 5+ participants)Hallway testing with 3 people for 15 minutesEarly concepts, internal tools, low-stakes features
Multi-week discovery research3 customer interviews in one week + support ticket reviewYou need directional insight, not statistical confidence
Design sprint (5 full days)Mini-sprint: 2-day version (map + sketch on day 1, prototype + test on day 2)Smaller problems, aligned team, good prior research
Full journey mapping workshopAssumption map on a whiteboard (30 min with the team, mark what you know vs. guess)When you need alignment more than research
Formal design critique (weekly, all designers)Async critique in Figma comments with one engineer and one PMSolo designer, no design peers available
Design system with governanceComponent inventory + shared Figma library with naming conventionsSmall product, 1-3 engineers

Building Relationships and Advocacy

As a solo or lean-team designer, your effectiveness depends on relationships more than process.

Earn a seat at the table:

  • Start by meeting the team's immediate needs — even grunt work builds goodwill
  • Show, don't tell. A quick prototype that surfaces a problem is worth more than a presentation about UX maturity
  • Connect UX work to metrics leadership already cares about (conversion, support tickets, retention)

Influence without authority:

  • Frame recommendations as experiments, not mandates: "Can we test this with 3 users before committing?"
  • Share research findings in the format your audience consumes — Slack summary for engineers, dashboard for leadership, annotated screenshots for PMs
  • Present findings with confidence but invite collaboration: "Here's what we learned — what does this mean for your area?"

Adapted Rituals for Solo/Lean Teams

Full RitualLean AlternativeCadence
Weekly design critiqueBi-weekly "design lunch" — show WIP to anyone interested, even non-designersEvery 2 weeks
Design-eng sync15-minute standup addition: "Here's what's coming in design this sprint"Per sprint
Formal design reviewShare a 1-page decision doc with the PM before a milestone. Async approvalPer milestone
Design QA sprintQA checklist in the PR template — designer reviews before mergePer story
RetroAdd one design-specific question to the team retro: "Where did design and implementation diverge?"Per sprint

Scaled-Down Handoff

When full specs aren't practical:

  1. Annotate in-context. Add behavior notes directly on Figma frames — don't create a separate spec document
  2. Prioritize what's ambiguous. Don't spec obvious things. Focus on states, transitions, and edge cases that engineers would have to guess
  3. Walk through once, document questions. A 15-minute walkthrough call + a list of engineer questions answered in Figma comments is often enough
  4. Build shared vocabulary early. If you and the engineer agree on component names and token usage, you need fewer words in every handoff

The Power of Small Wins

Track and communicate your impact, even informally:

  • "We tested the new flow with 3 users — 2 couldn't find the save button. We moved it, and support tickets for that flow dropped 40%"
  • "The assumption map revealed we had no data on how users think about billing. Two interviews later, we redesigned the billing page before engineering started"
  • "Adding a design QA step to PRs caught 12 visual bugs this quarter that would have shipped"

Small wins compound. Each one makes the next conversation about investing in design easier.


Common Mistakes

MistakeWhy It FailsBetter Approach
Skipping the sprint map dayTeam builds solutions to the wrong problemMonday's map and target selection is the most valuable day
Designing in isolationWork handed off doesn't match engineering realityDaily or weekly design-eng syncs
No design QA processVisual quality degrades over timeQA every story, not just every release
Over-documenting30-page spec nobody readsSpec what's ambiguous. Shared component system handles the rest
Under-documenting"You can see it in Figma"If an engineer has to ask, the spec was incomplete
Treating handoff as a one-time eventThrows work over the wallHandoff is a conversation that continues through implementation
No retrosSame mistakes repeatBi-weekly retros with specific action items and owners
Design critique as approval gateCritique becomes politicalCritique is for improving work, not for approving it

平台分布

Codex

34.39%
按下载量换算63

Claude

29.34%
按下载量换算53

Cursor

20.95%
按下载量换算38

Gemini CLI

10.41%
按下载量换算19

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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