Playwright Skill Safe

Use this skill to write and execute task-specific Playwright scripts safely.

Requirements

• Playwright installed
• Chromium installed
• Write scripts only to /tmp/playwright-test-*.js

Hard rules

• Always write scripts to /tmp/playwright-test-*.js
• Never write automation scripts into the skill folder or target project
• Always capture console errors, page errors, and request failures
• Always take a screenshot on failure
• Save trace artifacts on success/failure when possible
• Prefer page.getByRole(), page.getByLabel(), page.getByPlaceholder(), and page.getByText() over brittle selectors
• Use raw CSS selectors only as a last resort
• Default to headless Chromium unless visual debugging is useful
• Never auto-install packages at runtime

Standard workflow

1. Determine the target URL and exact browser task
2. Choose the right pattern:
   • deploy QA
   • responsive QA
   • form flow
   • login flow
   • visual verification
   • broken request/API smoke
3. Write a purpose-built script to /tmp/playwright-test-*.js
4. Run it with:

node "$SKILL_DIR/scripts/run_playwright_task.js" /tmp/playwright-test-*.js

5. Report:
   • PASS / FAIL
   • what was tested
   • screenshots saved
   • traces saved
   • console errors
   • request failures
   • HTTP 4xx/5xx errors
   • accessibility warnings

Built-in capabilities

• headed or headless mode via exported config
• viewport presets: desktop, laptop, tablet, mobile
• trace capture
• screenshot helpers
• network failure capture
• HTTP error capture
• accessibility smoke checks
• resilient locator helpers
• responsive sweep helper

Script contract

Generated scripts should export:

exports.config = {
  headed: false,
  slowMo: 0,
  // optional Playwright device name
  // device: 'iPhone 13'
};

exports.run = async ({ browser, context, page, result, helpers, chromium }) => {
  // your task here
  result.ok = true;
};

Recommended patterns

Deploy QA mode

• open URL
• wait for stable load
• capture title
• verify key heading or CTA
• take desktop screenshot
• run basic accessibility smoke

Responsive QA mode

• run desktop/tablet/mobile sweep
• capture each screenshot
• note obvious layout issues

Form flow mode

• use label/role/placeholder locators first
• fill realistic data
• submit
• verify success or validation state

Login flow mode

• navigate to login
• fill credentials if provided
• submit
• verify redirect or authenticated UI

Files

• Runner: scripts/run_playwright_task.js
• Helpers: scripts/helpers.js
• Reference examples: references/examples.md
• Task templates: references/templates.md

Resolve $SKILL_DIR to the directory containing this SKILL.md before running commands.