NYC

install-windows-3-11

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • COMMAND_EXECUTION (HIGH): The skill instructs the agent to perform administrative tasks including modifying files in /etc/nginx/, deleting the default site configuration (rm -f /etc/nginx/sites-enabled/default), and reloading the Nginx service via systemctl. These actions significantly alter the host system's web service configuration. Note: Severity is reduced to MEDIUM in the final verdict because these operations are central to the skill's primary purpose.
  • COMMAND_EXECUTION (MEDIUM): Includes instructions for raw network/socket manipulation using socat and nc to communicate with the QEMU Machine Protocol (QMP) and websockify. This could be leveraged for unauthorized internal service access if not restricted to localhost.
  • PROMPT_INJECTION (LOW): Surfaces for indirect prompt injection exist where the agent is prompted to read local system configuration files. Ingestion points: references/novnc_nginx_config.md reads /etc/nginx/nginx.conf. Boundary markers: Absent. Capability inventory: systemctl, rm, ln, qemu-system-i386. Sanitization: Absent.
  • COMMAND_EXECUTION (LOW): Generates a simple helper script (send-key.sh) at runtime to automate QMP commands. This is a low-risk template-based script generation for VM control.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:04 PM