The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides templates and methods for extracting text content from Windows UI elements (e.g., control titles, labels, and text values) of external applications. This creates an indirect prompt injection surface where an agent reading data from a malicious application could be influenced by instructions hidden within the UI text. Ingestion points: BasePage.get_text and BasePage.wait_until in SKILL.md. Boundary markers: There are no explicit delimiters or instructions to ignore embedded content when processing UI text. Capability inventory: The skill utilizes subprocess.Popen and Application.start for process management. Sanitization: No sanitization or validation of extracted UI strings is included in the provided code patterns.\n- [EXTERNAL_DOWNLOADS]: Recommends the installation of standard testing frameworks and utilities from official registries and project sites. This includes Python packages from PyPI and external binaries like FFmpeg from its official domain and Python via Microsoft's winget service.\n- [COMMAND_EXECUTION]: Utilizes subprocess.Popen to launch applications under test and manage recording utilities. It also provides automation scripts using PowerShell for setting up test environments and installing dependencies within a Windows Sandbox instance.

windows-desktop-e2e