19.1"Agentic Testing"

What they say: "We need agentic testing capabilities."

What they mean: Tests that can make decisions, not just follow scripts. The test system should be able to:

• Explore the application (not just verify known flows)
• Recover from unexpected states
• Generate new test cases based on what it observes
• Report findings in natural language

As of 2026 this is mainstream vocabulary, not hype — Playwright ships first-party test agents, and every major commercial platform markets "agentic execution."

What you should say: "Agentic testing means the test system has a reasoning loop — it observes the application state, decides what to do next, acts, and evaluates the result. In our framework, the coding agent is that loop. It reads test plans, drives the browser through the Playwright CLI skill, and reasons about whether results match expectations. Playwright's planner/generator/healer agents are the productized version of the same idea."