playwright-e2e

Hybrid E2E UI testing skill for AI coding agents (Qwen Code, Claude Code, Cursor).

Principle: Browser MCP / Playwright MCP = authoring & debugging layer. Playwright Test = execution layer (queue, retries, CI gate). Do NOT run the suite "through MCP" as the executor.

What it does

• Generate specs via live app exploration (MCP) to produce a plan in specs/.md and tests in tests/.spec.ts
• Execute as a queue with scripts/run_queue.sh (sequential, --rerun-failed)
• Produce Markdown artifacts per run via scripts/junit_to_md.py
• Wire into CI/CD via templates/github-actions-mac.yml (self-hosted macOS runner)

Structure

SKILL.md                      # skill logic + agent instructions
scripts/run_queue.sh          # queue execution + rerun-failed + artifacts
scripts/junit_to_md.py        # JUnit XML -> Markdown report
templates/playwright.config.ts# reporters (list/html/junit), retries, trace
templates/github-actions-mac.yml
references/best-practices.md  # 10 rules agents get wrong

Quick start

npm i -D @playwright/test
npx playwright install --with-deps
bash scripts/run_queue.sh --rerun-failed
# -> reports/run-<timestamp>.md

Portability

This is an Agent Skills-format skill. For Qwen Code:

cp -r . ~/.qwen/skills/playwright-e2e/

License: MIT