ventoy

Warn

Audited by Gen Agent Trust Hub on Oct 24, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The file gpt.md contains a hardcoded password (Password123!) within the autounattend.xml automation template provided for Windows 11 installations.
  • [PRIVILEGE_ESCALATION]: The skill frequently uses sudo to execute disk management tools, granting the agent administrative access to modify physical storage devices and system partitions.
  • [COMMAND_EXECUTION]: The skill instructions provide shell commands for direct disk manipulation using utilities such as dd, mkfs.ext4, parted, and Ventoy-specific shell scripts (Ventoy2Disk.sh).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data, specifically disk identifiers (e.g., /dev/sdb), which are interpolated into high-privilege commands.
  • Ingestion points: User-provided disk paths in SKILL.md, gemini.md, and gpt.md for formatting, updating, and installation commands.
  • Boundary markers: SKILL.md includes a mandatory safety directive instructing the agent to verify disks with lsblk before execution to prevent host storage overwrites.
  • Capability inventory: High-privilege disk partitioning (parted), filesystem creation (mkfs.ext4), block-level writing (dd), and installation scripts (Ventoy2Disk.sh) all executed via sudo.
  • Sanitization: No programmatic sanitization is present; the skill relies entirely on the agent following the natural language safety directive to verify disk identifiers.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 24, 2026, 04:33 AM