skills/malue-ai/dazee-small/winget/Gen Agent Trust Hub

winget

Warn

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Windows Package Manager (winget) to execute high-privilege operations including installing, upgrading, and uninstalling software on the host system. Commands like winget install and winget uninstall can significantly alter the security posture and stability of the operating system.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the download and execution of external software packages from the winget and Microsoft Store repositories. Although these are official distribution channels, they host third-party content that could be malicious if the package identifier is spoofed or manipulated.
  • [PROMPT_INJECTION]: The skill defines a mechanism for 'Interactions with other Skills' where it may automatically attempt to install software requested by a separate, potentially untrusted skill. This creates a surface for indirect prompt injection.
  • Ingestion points: Requests originating from other skills as described in the '与其他 Skill 联动' section.
  • Boundary markers: The instructions mandate 'HITL (Human-in-the-loop) confirmation' before any installation or uninstallation, which acts as a manual boundary marker.
  • Capability inventory: The skill possesses full access to winget installation, uninstallation, and upgrade commands.
  • Sanitization: There is no programmatic sanitization or validation of the application IDs against a known safe whitelist; it relies entirely on the user's manual review.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 6, 2026, 04:19 PM