browser-qa-testing

QA testing skill with integrated headless browser automation. Three tiers: Quick, Standard, Exhaustive. Persistent Chromium daemon, 60+ commands, @ref element selection.

Why

QA without a browser is guesswork for web apps. This skill runs a persistent headless Chromium daemon that starts in 3 seconds and responds in 100ms per command. It tests forms, validates responsive layouts, captures screenshots as evidence, and diffs page state before/after actions. The 6-phase test flow (Orient, Interact, Form, Responsive, Deep Inspect, Evidence) covers functional, visual, and security checks.

Install

git clone https://github.com/d-wwei/browser-qa-testing ~/.claude/skills/browser-qa-testing
cd ~/.claude/skills/browser-qa-testing
bash install.sh    # builds the browse binary (requires bun)

Or install automatically as part of Apex Forge.

Usage

/browser-qa-testing               # Standard tier (default)
/browser-qa-testing quick          # P0+P1 only, 30 seconds
/browser-qa-testing exhaustive     # Full audit including a11y, perf, network

Browser commands use the browse binary:

browse goto http://localhost:3000
browse snapshot -i                  # interactive elements with @refs
browse fill @e2 "test@example.com"
browse click @e3
browse snapshot -D                  # diff: what changed?
browse screenshot /tmp/evidence.png
browse console --errors

Tiers

Tier	Scope	Time
Quick	P0 + P1 findings	~30 seconds
Standard	P0 + P1 + P2	5-15 minutes
Exhaustive	P0-P3 + cosmetic + a11y + perf	15-30 minutes

Bug-Fix Loop

For each finding above the tier threshold: reproduce, write failing test, fix, verify with browser re-check, commit atomically.

Output Contract

output:
  health_score: number  # 0-100
  tier: enum [Quick, Standard, Exhaustive]
  test_results: list
    - category: string
      status: enum [PASS, FAIL, SKIP]
      evidence: string

License

MIT

---

browser-qa-testing

集成无头浏览器自动化的 QA 测试 skill。三级深度：Quick、Standard、Exhaustive。持久化 Chromium 守护进程，60+ 命令，@ref 元素选择。

为什么需要

Web 应用的 QA 不用浏览器就是盲测。这个 skill 运行持久化的无头 Chromium 守护进程，3 秒冷启动，每个命令 100ms 响应。测试表单、验证响应式布局、截图取证、diff 操作前后的页面状态。6 阶段测试流（定向、交互、表单、响应式、深度检查、证据收集）覆盖功能、视觉和安全检查。

安装

git clone https://github.com/d-wwei/browser-qa-testing ~/.claude/skills/browser-qa-testing
cd ~/.claude/skills/browser-qa-testing
bash install.sh    # 编译 browse 二进制（需要 bun）

或通过 Apex Forge 自动安装。

使用

/browser-qa-testing               # Standard 级（默认）
/browser-qa-testing quick          # 仅 P0+P1，30 秒
/browser-qa-testing exhaustive     # 全面审计，含无障碍、性能、网络

浏览器命令使用 browse 二进制：

browse goto http://localhost:3000
browse snapshot -i                  # 可交互元素 + @ref 标签
browse fill @e2 "test@example.com"
browse click @e3
browse snapshot -D                  # diff：什么变了？
browse screenshot /tmp/evidence.png
browse console --errors

分级

级别	范围	时间
Quick	P0 + P1	~30 秒
Standard	P0 + P1 + P2	5-15 分钟
Exhaustive	P0-P3 + 外观 + 无障碍 + 性能	15-30 分钟

Bug 修复循环

对每个超过当前级别阈值的发现：复现 → 写失败测试 → 修复 → 浏览器重新验证 → 原子提交。

输出契约

output:
  health_score: number  # 0-100
  tier: enum [Quick, Standard, Exhaustive]
  test_results: list
    - category: string
      status: enum [PASS, FAIL, SKIP]
      evidence: string

License

MIT