Agent Skills
› AjayIrkal23/agentic-mercy-10x
› qa-playwright
qa-playwright
GitHub基于 Playwright MCP 的系统化 QA 测试流程。涵盖启动服务、导航、快照、主路径与边界用例测试、多端响应式检查及控制台错误排查。发现 Bug 后执行修复循环并截图留存，确保功能完整且无错误。
触发场景
需要进行前端功能测试
需要验证 UI 响应式兼容性
需要系统性查找和修复 Bug
安装
npx skills add AjayIrkal23/agentic-mercy-10x --skill qa-playwright -g -y
SKILL.md
Frontmatter
{
"name": "qa-playwright",
"schema": 1,
"category": "general",
"surfaces": [
"general"
],
"triggers": {
"paths": [],
"intents": [
"general"
],
"keywords": [
"adapted",
"gstack",
"loop",
"mcp",
"playwright",
"systematic",
"testing"
]
},
"platforms": [
"linux",
"darwin",
"windows"
],
"token-cost": 237,
"description": "ALWAYS invoke for a systematic QA testing loop using Playwright MCP (adapted from GStack \/qa)."
}
QA with Playwright
Systematic QA loop: test → find bugs → fix → retest → commit.
Steps
- Start dev server if not running (
npm run devormake run) - Navigate to the feature:
mcp__playwright__browser_navigate - Snapshot the page:
mcp__playwright__browser_snapshot - Test golden path — click through primary user flow
- Test edge cases — empty states, error states, boundary inputs
- Test responsive — resize to 320px, 768px, 1440px:
mcp__playwright__browser_resize - Check console for errors:
mcp__playwright__browser_console_messages - Screenshot evidence:
mcp__playwright__browser_take_screenshot
Fix Loop
For each bug found:
- Document: what happened vs expected
- Fix the code
- Retest the specific scenario
- Screenshot the fix
- Continue testing
Completion
- All golden paths work
- No console errors
- Responsive at 3 breakpoints
- Screenshots captured as evidence
版本历史
- 581d130 当前 2026-07-19 09:13
