The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes macOS system commands and AppleScript to manage applications and system state (e.g., open_app, close_app, set_volume). It includes safeguards such as blocking dangerous keywords (rm, sudo) and validating application names against the /Applications/ directory.
[DATA_EXFILTRATION]: The skill possesses the capability to read sensitive user information through get_clipboard() and take_screenshot(). These operations expose local data to the agent's context, which is necessary for the skill's intended automation workflows.
[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection by ingesting untrusted data from the system clipboard and file system metadata via spotlight_search.
Ingestion points: System clipboard contents (get_clipboard), file search results and metadata (spotlight_search).
Boundary markers: No specific boundary markers or 'ignore' instructions are mentioned for the ingested data.
Capability inventory: File system access, application control, clipboard modification, and notification dispatching.
Sanitization: The skill documents that AppleScript strings are sanitized and shell=True is avoided to prevent direct command injection.

mac-control