Browser Automation Parity

Eliminate the gap between Browser Subagent mode (interactive, DOM-level reasoning) and Run mode (headless Python + Playwright) in Google Antigravity.

The core problem: automation that "passes" with subagent guidance silently fails or diverges when replayed as plain Python. This skill makes run mode observable, self-checking, and debuggable at step granularity.

When to Use

• Automation passes under Browser Subagent but fails (or silently diverges) in Python replay
• A missed click, paste, or navigation causes wrong results with unclear root cause
• You need CI-grade reproducibility with rich diagnostics
• Building any Playwright automation that should be production-grade

Key Principles

1. Every action is a checked contract