ai-video-generation

Fail

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides an installation command in the Quick Start section (curl -fsSL https://cli.inference.sh | sh) that fetches a remote script and executes it directly via the shell. This pattern allows for the execution of unverified code from a third-party domain.\n- [COMMAND_EXECUTION]: The skill instructs the agent to use the infsh tool via the Bash tool to process video generation requests. It interpolates user-supplied prompts and URLs directly into shell command arguments without providing sanitization or escaping mechanisms.\n
  • Ingestion points: User prompts and URLs provided to the infsh app run commands (SKILL.md).\n
  • Boundary markers: Absent; user input is placed inside JSON strings within shell command arguments.\n
  • Capability inventory: Access to the infsh binary via the Bash tool.\n
  • Sanitization: No instructions are provided for validating or escaping special characters in user-controlled inputs, which could be exploited for command injection.\n- [EXTERNAL_DOWNLOADS]: The installation process involves downloading binaries from dist.inference.sh.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 21, 2026, 04:55 PM