AI for QA / Quiz
Playwright MCP: Capability and Control
Read a browser agent trace, separate observation from proof, and choose controls for consequential actions.
- Difficulty
- Medium
- Format
- Quiz
- Points
- 150
- Estimate
- 8 min
// MISSION BRIEF
Your Mission
A browser-driving agent can move quickly through a web application, but its tool access does not make every conclusion reliable. Inspect a concrete MCP session and decide what the agent can establish, what remains outside its view, and where a tester must retain control.
// FIRST CONTACT
Battle teaser
The agent receives this fresh accessibility snapshot: - textbox "Email" [ref=e7] - textbox "Password" [ref=e9] - button "Sign in" [ref=e12] It must activate Sign in. Which next step best uses Playwright MCP's normal interaction model?
- AAssume the button was activated because the snapshot mentioned it
- BEstimate screen coordinates from a screenshot and click that pixel without checking the page snapshot
- CCall the browser click tool with the Sign in element and ref e12 from the fresh snapshot
- DInvent a database command that marks the user authenticated
Answers, scoring, hints, and the full battle stay sealed.
// SKILL TAGS
ai-framework-crossoversplaywright-mcpguardrailsai-for-qa