peekaboo

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the peekaboo binary to drive macOS UI interactions, including simulating complex inputs like drag, swipe, and hotkey sequences, as well as managing application life cycles.\n- [EXTERNAL_DOWNLOADS]: The skill's installation metadata specifies fetching the peekaboo formula from the steipete/tap/peekaboo Homebrew tap, which is necessary for the tool's functionality.\n- [DATA_EXFILTRATION]: The skill grants access to sensitive system data through commands that can read the clipboard (clipboard) and capture images of the screen or frontmost windows (image, capture, see).\n- [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection because it allows the agent to ingest and analyze arbitrary text from the macOS user interface.\n
  • Ingestion points: Data from the UI is ingested through the see and capture commands.\n
  • Boundary markers: The skill does not define specific delimiters or instructions to prevent the agent from obeying commands embedded in the processed UI text.\n
  • Capability inventory: The agent possesses powerful capabilities to interact with the system based on its interpretation of the UI content.\n
  • Sanitization: No explicit sanitization or filtering of text captured from application windows is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 01:08 AM