19.5"Test Observability"

What they say: "What's your observability strategy for tests?"

What they mean: Can you see what's happening inside your tests? Can you debug failures efficiently?

What you should say: "We capture the command log (what the agent did), snapshots and screenshots (what the page was), and the full Playwright trace — which since 1.60 includes network capture — as the single richest artifact: every action with before/after state. On failure, we add the agent's reasoning about why it thinks the test failed. This gives both human reviewers and AI agents everything they need to diagnose issues, and it doubles as the compliance answer to 'show me what the AI did.'"