use-computer-mcp
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured guidelines for an AI agent to use computer-interaction tools. All tool calls and parameters mentioned (such as
computer_get_app_state,computer_click, and profile switching) are consistent with the skill's stated purpose of facilitating desktop application automation. - [INDIRECT_PROMPT_INJECTION]: The skill utilizes tools that ingest external UI state and web content, which presents a surface for indirect prompt injection. This is a characteristic of the tool's function rather than a flaw in the skill instructions.
- Ingestion points: The agent retrieves UI accessibility trees and application state via
computer_get_app_stateas described inSKILL.md. - Boundary markers: The instructions do not explicitly mandate delimiters or provide specific instructions to the agent to disregard instructions found within the processed UI content.
- Capability inventory: The agent has permissions to perform system-level interactions, including clicking, typing, and setting values (
computer_click,computer_type_text,computer_set_value) across the host environment. - Sanitization: No specific filtering or sanitization of the incoming UI state data is performed before the agent uses it to decide subsequent actions.
Audit Metadata