The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the ctypes library to interface with low-level Win32 APIs in user32.dll and kernel32.dll, enabling operations such as input simulation (SendInput), window management (SetForegroundWindow), and process interaction. These capabilities are necessary for the skill's primary function and are documented alongside defensive patterns including input rate limiting and application blocklists.
[INDIRECT_PROMPT_INJECTION]: The skill implements patterns for element discovery and property retrieval that read text from external application UI elements, creating an attack surface for instructions embedded in third-party software.
Ingestion points: references/advanced-patterns.md (ElementDiscovery.find_element_path), SKILL.md (SecureUIAutomation.find_element).
Boundary markers: The skill emphasizes process identity validation and integrity checks but lacks explicit instructions for using delimiters or boundary markers when processing raw UI text strings.
Capability inventory: System-level interaction capabilities including Win32 input simulation, focus management, and COM automation via comtypes and UIAutomationClient.dll.
Sanitization: Includes a ProcessValidator class for verifying executable integrity, a SafeInputSimulator to block dangerous system key combinations, and a sanitize_property_value function in references/security-examples.md to remove control characters from retrieved data.

windows-ui-automation