ai-voice-cloning

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • REMOTE_CODE_EXECUTION (CRITICAL): The skill executes remote code from an untrusted source by piping a curl response to the shell (curl -fsSL https://cli.inference.sh | sh). This allows for arbitrary code execution with the user's privileges without prior verification of the script content.
  • EXTERNAL_DOWNLOADS (CRITICAL): The installation script and subsequent binary dependencies are fetched from cli.inference.sh, which is not a verified or trusted domain according to the security policy.
  • PROMPT_INJECTION (LOW): The skill is susceptible to indirect prompt injection because it interpolates untrusted user text into command-line arguments without sanitization or boundary markers.
  • Ingestion points: SKILL.md (various lines) where text for voice synthesis is passed to the --input flag of the infsh CLI.
  • Boundary markers: Absent; no specific instructions or delimiters are used to ensure the agent treats input as literal data.
  • Capability inventory: Shell execution of the infsh binary via the Bash tool.
  • Sanitization: Absent; no input validation is performed.
  • COMMAND_EXECUTION (MEDIUM): The skill requires the Bash(infsh *) tool, granting it the capability to execute a custom binary that performs remote network operations and processes data on an external platform.
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
CRITICAL
Analyzed
Feb 17, 2026, 06:45 PM