iphone-mirroring

Warn

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill includes a bash wrapper script (scripts/iphone-mirror) that automatically compiles a Swift source file (scripts/iphone-mirror.swift) into an executable binary using swiftc on the first run. The resulting binary is cached in the user library and executed for subsequent operations.
  • [COMMAND_EXECUTION]: The skill executes various system commands, including /usr/sbin/screencapture to capture the content of the iPhone Mirroring window. It also interacts with macOS Accessibility APIs to simulate user input events such as clicks, scrolls, and swipes, and to manipulate application menus.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process visual data (screenshots) from a real iPhone. This creates an attack surface where malicious instructions or content displayed within a mobile application (such as a post or a message) could attempt to influence the agent's actions. The skill documentation includes rules for manual verification and a default read-only policy to help mitigate this risk.
  • [PRIVILEGE_ESCALATION]: To function, the skill requires high-level system permissions, including Accessibility and Screen & System Audio Recording. While the skill does not attempt to bypass security controls, it guides the user through the process of manually granting these significant privileges to the agent's host application in System Settings.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 11, 2026, 02:02 AM