What problem does it solve? Manually navigating websites, filling forms, and extracting page data is repetitive and error-prone, and traditional browser automation floods AI context with full DOM output. This Skill drives a real browser through the agent-browser CLI using compact element refs, cutting token usage per interaction while handling navigation, input, capture, and session state. ## Core Features & Use Cases - Ref-Based Interaction: Snapshot pages to get compact @refs, then click, fill, select, drag, and upload against those refs instead of brittle CSS selectors. - Capture & Extraction: Take full-page screenshots, save pages as PDF, record session videos, and extract text, HTML, attributes, and computed styles. - Sessions & Authentication: Run isolated parallel sessions, save and restore login state, handle OAuth and 2FA flows, and route traffic through proxies. - Use Case: Log into a web app once, save the auth state to a file, then reuse it across runs to scrape dashboard data and capture screenshots without repeating the login flow. ## Quick Start Ask the agent to open a website with agent-browser, take an interactive snapshot, fill in the login form, and save a screenshot of the result.
