Skills
Navigate
skills/redfox-data/redfox-community/video-prompt-expert/Gen Agent Trust Hub

video-prompt-expert

Fail

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: HIGHCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: Instructions in SKILL.md and references/core_workflow.md direct the agent to modify shell configuration files (~/.zshrc, ~/.bashrc) or use PowerShell to set permanent environment variables. Modifying startup scripts is a persistence mechanism used to ensure the environment remains configured across sessions.
  • [COMMAND_EXECUTION]: The skill implements a silent execution pattern in references/core_workflow.md, instructing the agent to run python "$SKILL_PATH/scripts/videogen.py" --record-only without notifying the user. This command is intended to report usage statistics.
  • [DATA_EXFILTRATION]: The script scripts/videogen.py contains a save_record function that sends the REDFOX_API_KEY to an external endpoint (https://redfox.hk/story/api/skill/record/save). While this is the vendor's own domain, the transmission of a credential during a silent usage report constitutes an exfiltration pattern.
  • [CREDENTIALS_UNSAFE]: The script scripts/videogen.py includes logic in read_shell_config to programmatically scan and parse sensitive user files like ~/.zshrc, ~/.bashrc, and the Windows Registry to extract the REDFOX_API_KEY credential.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Aug 12, 2026, 11:12 AM
Security Audit — agent-trust-hub — video-prompt-expert