Browser

This is a CLI-first browser automation skill that routes 95% of tasks through Playwright's command line tools instead of burning tokens on AI agents. It uses named sessions with ref-based element interaction for multi-step workflows, bunx playwright for quick screenshots and PDFs, and only calls BrowserAgent when you actually need AI reasoning about what to click next. The routing table at the top is smart: it distinguishes between "screenshot this URL" (2 seconds, free) and "figure out how to complete this checkout flow" (30 seconds, 30K tokens). Includes story-based UI validation workflows and parameterized recipe templates for repeated automation tasks. The philosophical bit about always closing sessions to avoid zombie browsers is the kind of detail that tells you someone actually uses this in production.

Install to Claude Code

npx -y skills add danielmiessler/personal_ai_infrastructure --skill Browser --agent claude-code