browseros-agent-harnesslisted

# BrowserOS & BrowserOS Neo Agent Skill Use this skill whenever executing web browser automation tasks, account login operations, DOM manipulation, or browser-based AI agent workflows across Claude Code, Gemini CLI, Cursor, Codex, and Antigravity. ## Core Capabilities & Directives ### 1. Dedicated Agent Browser (`BrowserOS Neo`) - **Protocol**: Leverages `browseros-neo` MCP server tools (`tabs`, `navigate`, `snapshot`, `act`, `download`, `upload`, `read`, `grep`, `screenshot`, `pdf`, `wait`, `evaluate`, `run`). - **Profile**: Uses Igor's active persistent local profile with pre-authenticated logins across Skool, GitHub, Stripe, Linear, and email portals. - **Privacy-First**: Operates locally with zero mandatory cloud relays. ### 2. Multi-LLM Driver Support - Supports 13 AI providers including local Ollama (PagedAttention harness), OpenRouter (500+ models), Moonshot Kimi K2.5, Google Gemini, and Anthropic Claude. ### 3. Automated Web Actions - Execute `node tools/browseros-agent-harness.js --status` to inspect active MCP tools and persistent profile health. - Execute `node tools/browseros-agent-harness.js --navigate <URL>` for zero-human-intervention page navigation and snapshotting. ### 4. Built-in App Integrations (MCP) - Connects browser DOM state to 40+ app integrations including Gmail, Slack, Notion, and Google Calendar.