ai-marketing-videos

Fail

Audited by Gen Agent Trust Hub on Feb 24, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructions include a procedure to install the infsh CLI by downloading a script from https://cli.inference.sh and piping it directly into a shell. This is a standard installation method for the service.
  • [EXTERNAL_DOWNLOADS]: The skill fetches and interacts with various external AI models and sub-skills from the inference.sh ecosystem, including components for video, image, and voice synthesis.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool restricted to the infsh command prefix to run video generation workflows, merge media files, and manage application states.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it processes user-supplied prompts for marketing scenes and script content.
  • Ingestion points: User-provided text for video descriptions, product features, and voiceover scripts used in the infsh app run examples.
  • Boundary markers: None identified; user input is interpolated as string arguments within JSON payloads in CLI commands.
  • Capability inventory: The skill has permissions to execute the infsh utility, which communicates with remote APIs and saves data to local files (e.g., hook.json).
  • Sanitization: There is no explicit evidence of input sanitization or filtering to prevent malicious instructions within user-provided strings from affecting the AI model's behavior.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 24, 2026, 07:50 AM