playwright-ci-caching_skill

Overview

This skill caches Playwright browser binaries in CI/CD pipelines to eliminate the 1–2 minute download overhead on every build. It extracts the Playwright package version to build a cache key, stores platform-appropriate cache paths, and only installs browsers when the cache is missed. Works with GitHub Actions and Azure DevOps and includes a helper PowerShell script to discover and run the Playwright CLI.

How this skill works

The pipeline reads Microsoft.Playwright version from Directory.Packages.props and writes it to an environment variable. The cache step uses runner OS plus that version as the cache key and stores the OS-specific Playwright browser cache folders. A conditional install step runs the Playwright CLI only if the cache miss occurs. The version in the key ensures automatic invalidation when Playwright is upgraded.

When to use it

• You run Playwright E2E tests in CI and want faster builds
• Builds are slowed by recurring ~400MB browser downloads
• You want automatic cache invalidation on Playwright upgrades
• You use GitHub Actions or Azure DevOps pipelines
• You can run PowerShell on CI agents

Best practices

• Extract Playwright version from Directory.Packages.props to form the cache key
• Include runner OS plus version in the key to avoid cross-OS conflicts
• Cache all OS paths in multi-OS workflows to ensure hits on each runner
• Run dotnet restore/build before attempting to find Microsoft.Playwright.dll
• Only run playwright install when cache-hit is false to skip redundant downloads

Example use cases

• GitHub Actions workflow that caches ~/.cache/ms-playwright with key ${{ runner.os }}-playwright-${{ env.PlaywrightVersion }}
• Multi-OS GitHub workflow listing Linux, macOS, and Windows cache paths for a shared cache step
• Azure DevOps pipeline using Cache@2 with key 'playwright | $(Agent.OS) | $(PlaywrightVersion)'
• Using build/playwright.ps1 to discover Microsoft.Playwright.dll and invoke the Playwright CLI after dotnet build
• Faster PR feedback loops by avoiding network-bound browser downloads