computer-use
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill enables the agent to interact with and process data from arbitrary external user interfaces (web browsers and desktop applications), which introduces an indirect prompt injection attack surface.
- Ingestion points: The skill reads application interfaces using accessibility data and visual inspection, as noted in SKILL.md.
- Boundary markers: The instructions do not define specific delimiters or boundary markers for content read from the UI.
- Capability inventory: The agent is authorized to perform clicks, typing, and navigation within the user's environment.
- Sanitization: The skill mandates safety checkpoints, instructing the agent to 'Ask before new permissions, charges, destructive changes, or sensitive account actions.'
- [COMMAND_EXECUTION]: The skill provides the agent with capabilities to drive a graphical user interface, effectively allowing the execution of system commands and interactions with application settings.
- Evidence: The skill instructions allow the agent to 'operate real apps and browsers: read interfaces, click, type, and verify results.'
Audit Metadata