agent-desktop-macos
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- System Permission Requirements: The skill facilitates the management of high-privilege macOS permissions, specifically Accessibility, Screen Recording, and Automation. These permissions allow the underlying tool to monitor and control the desktop environment, which is the primary intended function of the skill. Users are directed to grant these through standard macOS System Settings.
- Indirect Prompt Injection Surface: By interacting with the macOS UI and Notification Center, the skill creates a surface for indirect prompt injection where untrusted data could influence agent behavior. * Ingestion points: Untrusted data enters the agent's context through UI tree snapshots, window titles, and notification banners via the
snapshotandlist-notificationscommands described in SKILL.md. * Boundary markers: The instructions do not define specific delimiters or guidelines for the agent to distinguish between observed UI text and its own instructions. * Capability inventory: The skill provides the agent with the ability to perform system-wide actions including mouse clicks, keyboard input, and application management (launchandclose-app). * Sanitization: There is no mention of sanitizing or filtering the content of UI elements or notifications before they are processed by the agent.
Audit Metadata