e2e-testing_skill

Overview

This skill provides comprehensive end-to-end testing for web applications using Playwright MCP. It observes and reports issues, discovers alternative paths, and tracks test history to identify flaky or regressing behavior. Use it to set up test regimes, run automated exploratory tests, and generate human- and machine-readable reports.

How this skill works

The skill operates in three modes: setup (discover flows and write a test regime), run (execute scenarios sequentially with full evidence capture), and report (produce actionable summaries and JSON output). During runs it records screenshots, DOM snapshots, network and console logs, evaluates success with configurable flexibility rules, and updates a history file to flag regressions and flaky scenarios.

When to use it

• When you need to create or update an E2E test regime for a web app
• When you want to run deterministic, evidence-backed E2E test runs
• When exploring the app to discover undocumented or alternative paths
• When investigating flaky tests or regressions across test history
• When you need both human-readable and machine-readable test reports

Best practices

• Confirm Playwright MCP is available before starting; provide base URL and key workflows
• Start Setup mode with URL, purpose, and any docs to maximize discovery quality
• Keep tests deterministic: always start runs from scenario step 1 and avoid cached state
• Treat missing features or incomplete verifications as failures, not skips—only environment outages justify SKIPPED
• Capture full evidence per step (screenshot, DOM, network, console, accessibility) for reproducibility
• Use discovery_cap to limit exploratory paths and add suggested variations for flaky scenarios

Example use cases

• Set up an initial regime for a newly deployed web app by exploring key user journeys
• Run nightly E2E suites that capture screenshots and DOM snapshots for every step
• Analyze recent runs to find flaky scenarios and generate suggested test variations
• Investigate a blocking failure in checkout with a machine-readable report that includes reproduction Playwright commands
• Discover undocumented guest checkout or alternate navigation paths and add them to the regime