The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of several external components from public registries to function:
Node.js packages @hypothesi/tauri-mcp-server and @hypothesi/tauri-mcp-cli via NPM.
The Rust plugin tauri-plugin-mcp-bridge via Cargo.
The utility install-mcp via npx for automated configuration.
[DYNAMIC_EXECUTION]: The webview_execute_js tool provides the capability to execute arbitrary JavaScript code within the context of the application's webview. This is a powerful feature intended for UI diagnostics and automated testing.
[COMMAND_EXECUTION]: The skill includes tools such as ipc_execute_command, which allows the AI to invoke backend Tauri commands, and manage_window for system-level window manipulations. The webview_interact and webview_keyboard tools further permit programmatic control over the application's user interface.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a data ingestion surface where the agent processes content from the external application, creating a potential for indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through logs (console, Android logcat, iOS logs), captured IPC traffic via ipc_get_captured, and structural application data via webview_dom_snapshot.
Boundary markers: The skill documentation does not mention the use of boundary markers or instructions to ignore embedded commands within ingested logs or IPC payloads.
Capability inventory: The agent possesses high-impact capabilities, including arbitrary JavaScript execution (webview_execute_js), backend command execution (ipc_execute_command), and simulated user input (webview_keyboard, webview_interact).
Sanitization: There is no indication of data sanitization or filtering before the external application data is processed by the AI assistant.

tauri-mcp-server-development