peekaboo

Warn

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of a third-party CLI tool from an external Homebrew tap via brew install steipete/tap/peekaboo. This pattern installs binaries from a non-standard repository that is not included in the platform's trusted vendor list.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute arbitrary shell commands for installation (brew install) and UI automation (peekaboo commands). This allows the agent to interact with the underlying operating system to capture screens and simulate user input.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the accessibility trees of running macOS applications, which can be manipulated by malicious applications or websites.
  • Ingestion points: Data retrieved via peekaboo inspect which reads labels, names, and values from the macOS accessibility API for running applications (SKILL.md).
  • Boundary markers: Absent. There are no instructions to the agent to ignore or sanitize instructions found within UI element labels.
  • Capability inventory: The skill has access to Bash (command execution) and Write (file system modification) across its operational flow.
  • Sanitization: Absent. UI element content is processed directly to identify targets for clicking or typing without validation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 7, 2026, 02:13 AM