Playwright CI Reliability AI Skill Guide

Overview & Engine Architecture

This skill complements @playwright with pipeline-first patterns. CI runs browsers headlessly under time pressure; flakes usually come from shared state, network races, or insufficient isolation - not "bad CI luck." Agents configure retries + trace-on-first-retry, shard large suites, merge blob reports, and upload HTML/trace artifacts for every failing job.

CI job matrix (shards)
        |
 playwright test --shard=i/n
        |
 blob report per shard
        |
 merge-reports -> HTML
        |
 upload traces / videos on failure

When to use this skill