Opens a traced browser session and records every action as a ruvector trajectory inside an RVF cognitive container. You invoke this whenever you need a browser interaction that isn't throwaway, because downstream skills like browser-extract and browser-form-fill compose on top of it. The lifecycle is still bash-driven until the native MCP tool ships, so you're managing session IDs manually in YYYYMMDD-HHMMSS format and calling ruvector hooks before and after each step. The trajectory data gets indexed in AgentDB under browser-sessions for later replay or analysis. It's a primitive, so redaction and content policy are handled upstream by the skills that actually read what you navigated to.
npx skills add https://github.com/ruvnet/ruflo --skill browser-record