video-gen

Fail

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill automatically performs a global installation of a Node.js package from an unverified scope (@marswave/listenhub-cli) using npm install -g if the command is not found. This occurs without requesting user permission.\n- [REMOTE_CODE_EXECUTION]: Following the installation of the untrusted package, the skill proceeds to execute commands from that package (e.g., listenhub video create) to perform video generation tasks.\n- [COMMAND_EXECUTION]: The skill makes extensive use of shell command execution, including ffprobe for metadata extraction and curl for downloading external video files to the local file system.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection.\n
  • Ingestion points: User video descriptions, reference image URLs, and source video URLs (SKILL.md Step 1, 3a-3d).\n
  • Boundary markers: Absent. The skill uses shell quoting but lacks explicit delimiters or instructions to ignore embedded commands.\n
  • Capability inventory: Execute CLI commands via listenhub and write to the local filesystem via curl (SKILL.md Step 5 and Submit).\n
  • Sanitization: Absent. There is no validation of the content of user-provided strings or external assets before they are used in sensitive operations.\n- [CREDENTIALS_UNSAFE]: The skill programmatically interacts with sensitive authentication information, such as OpenAPI keys, by querying the CLI's configuration and status (SKILL.md Step -1). It manages these credentials to determine the appropriate command prefix for execution.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Aug 30, 2026, 08:21 PM
Security Audit — agent-trust-hub — video-gen