Tauri MCP Test Runner Overview Run VMark end-to-end UI flows using the Tauri MCP tools and summarize results. Workflow Confirm the desktop app is running. If not, ask the user to start it. Start MCP session via mcp__tauri__tauri_driver_session (action: start , port: 9223 or omit for default). List windows ( mcp__tauri__tauri_manage_window action: list ) and pick the target window. Execute test flows using the patterns in references/flows.md . Capture logs when needed ( mcp__tauri__tauri_read_logs ) and stop the session when done. Report results: passed/failed flows, errors, repro steps, and any follow-up fixes. Key Tools Tauri MCP Tools (tauri-mcp server) tauri_driver_session - Start/stop driver session tauri_webview_screenshot - Capture current state tauri_webview_keyboard - Send keystrokes (Enter, Escape, shortcuts) tauri_webview_interact - Click, focus elements tauri_webview_find_element - Inspect DOM state VMark MCP Tools (vmark server) document_insert_at_cursor - Insert text (appears as AI suggestion) cursor_set_position - Move cursor to position cursor_get_context - Get surrounding text context selection_set / selection_get - Manage text selection format_toggle - Toggle formatting (bold, italic, etc.) Critical Pattern: AI Content Acceptance When using vmark MCP to insert content, it appears as an AI suggestion that must be accepted with Enter via tauri-mcp: 1. vmark: document_insert_at_cursor("text") 2. tauri: keyboard(action: "press", key: "Enter") # Accept 3. tauri: keyboard(action: "press", key: "Escape") # Dismiss follow-up This pattern is required for all content insertion operations. References Load references/flows.md for detailed flow patterns and examples.
tauri-mcp-test-runner
安装
npx skills add https://github.com/xiaolai/vmark --skill tauri-mcp-test-runner