做什么
Claude Code的AI工作流程专家小组,负责CEO审查、工程规划、代码审查、运输、质量保证测试和浏览器自动化
技能库 智客分类:Agent 工作流 gstack-workflow-assistant
Claude Code的AI工作流程专家小组,负责CEO审查、工程规划、代码审查、运输、质量保证测试和浏览器自动化
官方网址:skills.sh
先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。
Claude Code的AI工作流程专家小组,负责CEO审查、工程规划、代码审查、运输、质量保证测试和浏览器自动化
官方 description 未单独写出 Use when。按规范,代理会在用户任务与这段 description 的关键词匹配时激活本技能。
按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:gstack Workflow Assistant、What It Does、Installation、Requirements、Global Installation、Project Installation。 其中含规范建议的小节:分步指令。
文件分析:这是一份仅含 SKILL.md 的指令型技能,代理激活后整份正文进入上下文。
Team of specialist AI workflows for Claude Code with CEO review, engineering planning, code review, shipping, QA testing, and browser automation
gstack Workflow AssistantWhat It DoesInstallationRequirementsGlobal InstallationProject InstallationgstackCore CommandsPlanning WorkflowsCode Quality WorkflowsQA and Testing WorkflowsTeam Workflows
来源分类:skills.sh agent-skill
namegstack-workflow-assistantdescription具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗
先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。
复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。
把 Agent Skill「gstack-workflow-assistant」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-c5716b1bedf0c699-Gstack-%E5%B7%A5%E4%BD%9C%E6%B5%81%E7%A8%8B%E5%8A%A9%E7%90%86.html 请存为 .cursor/skills/gstack-workflow-assistant/SKILL.md 或 .claude/skills/gstack-workflow-assistant/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。
需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。
npx skills add 'https://github.com/reason-machines/trending-skills' --list
npx skills add 'https://github.com/reason-machines/trending-skills' --skill 'gstack-workflow-assistant'
CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。
name: gstack-workflow-assistant description: Team of specialist AI workflows for Claude Code with CEO review, engineering planning, code review, shipping, QA testing, and browser automation triggers:
Skill by ara.so — Daily 2026 Skills collection
Transform Claude Code from a generic assistant into a team of specialists you can summon on demand. Eight opinionated workflow skills that act as CEO, Engineering Manager, Release Manager, and QA Engineer with slash commands for planning, review, shipping, and testing.
gstack provides specialized AI personas and workflows:
git clone https://github.com/garrytan/gstack.git ~/.claude/skills/gstack
cd ~/.claude/skills/gstack
./setup
cp -Rf ~/.claude/skills/gstack .claude/skills/gstack
rm -rf .claude/skills/gstack/.git
cd .claude/skills/gstack
./setup
Add to your CLAUDE.md:
## gstack
Use /browse skill from gstack for all web browsing. Never use mcp__claude-in-chrome__* tools.
Available skills:
- /plan-ceo-review - Product strategy review
- /plan-eng-review - Technical architecture planning
- /review - Thorough code review
- /ship - One-command branch shipping
- /browse - Browser automation and testing
- /qa - Systematic QA testing
- /setup-browser-cookies - Session management
- /retro - Engineering retrospectives
// Start with feature description, then review strategy
You: I want to add seller photo upload to the listing app
You: /plan-ceo-review
// AI responds as CEO: challenges assumptions, finds bigger opportunity
// "Photo upload" → "AI-powered listing creation from photos"
You: /plan-eng-review
// AI responds as tech lead:
// - Architecture diagrams
// - State machines
// - Async job boundaries
// - Failure modes
// - Test matrices
You: /review
// Paranoid staff engineer review:
// - Race conditions
// - Trust boundaries
// - Missing error handling
// - Production failure modes
You: /ship
// Automated release process:
// 1. Sync main branch
// 2. Run test suite
// 3. Resolve review comments
// 4. Push branch
// 5. Open pull request
You: /browse https://myapp.com
// AI navigates your app:
// - Takes screenshots
// - Clicks through workflows
// - Identifies breakage
// - Tests responsive design
You: /qa
// Branch-aware testing:
// - Analyzes git diff
// - Identifies affected pages
// - Tests localhost:3000
// - Full exploration mode
// - Regression testing
You: /qa https://staging.myapp.com --quick
// Quick smoke test: 5 pages in 30 seconds
You: /setup-browser-cookies staging.myapp.com
// Imports cookies from real browser (Chrome, Arc, Brave, Edge)
// Enables testing authenticated pages without manual login
You: /retro
// Engineering manager-style retro:
// - Analyzes git history
// - Per-person contributions
// - Growth opportunities
// - Team dynamics
// - Saves to .context/retros/
gstack creates isolated browser instances in .gstack/ directory:
// Automatic browser configuration
{
userDataDir: '.gstack/browser-data',
headless: false, // for debugging
viewport: { width: 1280, height: 720 },
timeout: 30000
}
your-project/
├── .claude/
│ └── skills/
│ └── gstack/
│ ├── skills/ # Workflow prompts
│ ├── browse/ # Browser automation
│ └── package.json
├── .gstack/ # Browser data (gitignored)
│ ├── browser-data/
│ └── screenshots/
└── .context/
└── retros/ # Retrospective history
Use Conductor for parallel sessions:
// Session 1: Feature development
You: /plan-ceo-review
You: /plan-eng-review
// [implement feature]
// Session 2: Code review
You: /review
// [fix issues]
// Session 3: QA testing
You: /qa --full
// Session 4: Release
You: /ship
# .github/workflows/gstack-qa.yml
name: gstack QA
on: [pull_request]
jobs:
qa:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- run: cd .claude/skills/gstack && ./setup
- run: echo "/qa --regression" | claude-code
# Add to team docs
## Development Workflow
1. `/plan-ceo-review` - Validate product direction
2. `/plan-eng-review` - Lock architecture
3. Implement feature
4. `/review` - Paranoid code review
5. `/qa` - Test branch thoroughly
6. `/ship` - One-command release
7. `/qa staging.app.com` - Staging verification
// Feature-specific testing
You: /qa --focus=checkout-flow
You: /qa --mobile-only
You: /qa --accessibility
// Performance testing
You: /browse https://app.com
// Then: "Run lighthouse audit on this page"
// Test in multiple browsers
You: /setup-browser-cookies app.com
You: /browse app.com --browser=chrome
// [switch session]
You: /browse app.com --browser=firefox
// Before major changes
You: /qa --baseline
// [make changes]
You: /qa --compare-baseline
// AI identifies visual/functional regressions
# Rebuild browser automation binary
cd .claude/skills/gstack
rm -rf browse/dist
./setup
# Fix executable permissions
cd .claude/skills/gstack
chmod +x setup browse/dist/browse
# Reinstall Node dependencies
cd .claude/skills/gstack
rm -rf node_modules
bun install
// Debug browser issues
You: /browse --debug
// Runs browser in non-headless mode for inspection
// Clear browser data
rm -rf .gstack/browser-data
// Re-register skills
cd .claude/skills/gstack
./setup
// Verify skills are available
You: /plan-ceo-review --help
The skills work by providing specialized prompts and browser automation tools that transform Claude Code into domain experts. Each workflow enforces specific cognitive modes - CEO strategic thinking, engineering rigor, paranoid review, or systematic testing - instead of generic assistance.
Agent 工作流
创造有效技能指南。 当用户想创造出新的技能(或更新现有的技能),以专业知识,工作流程,或工具集成来扩展克洛德的能力时,应该使用这种技能.
Agent 工作流
使用ClawdHub CLI搜索,安装,更新并发布从taladhub.com的代理技能. 需要获取苍蝇上的新技能时使用,将安装的技能同步到最新版本或特定版本,或者发布 npm-instainddhub CLI 的新/更新的技能文件夹.
Agent 工作流
管弦乐团多代理团队,任务设定周期,交接协议,审查工作流程. 使用时间: (1)建立2+特派员队伍,具有不同专业,(2)确定任务路线和生命周期(收录框_ spec_建设_审查_完成),(3)在特派员之间制定交接协议,(4)建立审查和质量关口,(5)管理特派员之间的交流和文物共享.
Agent 工作流
Spec-first,TDD,子代理驱动的软件开发工作流程. 当:(1)构建任何新功能或应用——触发脑暴_计划_子代理执行回路,(2)调试出一个bug或测试失败——触发系统性的根起过程,(3)用户说"让我们构建","帮助我计划","我想添加X",或"这个被打破",(4)完成一个功…