web-testing

Scope

Covers driving a real browser against a running application and turning that into a suite that can be trusted: locator strategy and strict mode; the actionability checks behind auto-waiting and why a fixed wait is a defect rather than a slowdown; retrying web-first assertions; authentication state captured once and reused; network interception, HAR replay and clock control; traces, screenshots, videos and the post-mortem read of a failure; flakiness as a diagnosis problem with a retry budget; parallelism, isolation, locks, sharding and report merging; browsers in CI containers; and the trust boundary that applies when an agent is the one driving the browser.

Facts here are written against Playwright 1.63.0 with Chromium 153; rules that need a newer release than the project's carry a gate such as (Playwright 1.63+).

Not covered: