peekaboo
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates extensive interaction with the macOS operating system via the
peekaboocommand-line interface. This includes launching applications, managing windows, simulating keystrokes (hotkey,type), and manipulating the mouse (click,drag).- [INDIRECT_PROMPT_INJECTION]: The skill provides tools to ingest untrusted data from the user's environment which could be used to influence agent behavior through hidden instructions in UI elements or the clipboard. - Ingestion points: Screen content captured via
peekaboo seeandpeekaboo capture, and text from the system clipboard viapeekaboo clipboardare all potential vectors for malicious instructions. - Boundary markers: The instructions do not define boundaries or provide warnings to the agent to ignore instructions found within captured UI elements or clipboard data.
- Capability inventory: The agent possesses broad system capabilities through the
peekabooCLI, including clicking UI elements, typing text, and launching arbitrary applications. - Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the screen or clipboard before it is processed by the agent.
Audit Metadata