qwen-image-2-pro

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes the belt CLI tool to authenticate and run image generation tasks. The shell environment is limited to commands starting with the belt prefix through the allowed-tools configuration.\n- [EXTERNAL_DOWNLOADS]: The skill references documentation and setup scripts from the inference-sh GitHub repository. These downloads are necessary for the skill's integration with the inference.sh platform.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided strings that are passed to an external tool, creating a surface for potential injection.\n
  • Ingestion points: User-provided prompts and image URIs enter the agent context via input parameters.\n
  • Boundary markers: The skill uses JSON formatting for CLI inputs, but does not provide explicit instructions to separate control logic from user data within the prompt strings.\n
  • Capability inventory: The skill possesses the capability to execute shell commands and perform network operations to the inference service.\n
  • Sanitization: There is no evidence of prompt validation or sanitization before the input is passed to the belt command.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 02:22 PM
Security Audit — agent-trust-hub — qwen-image-2-pro