qwen-image-2
Pass
Audited by Gen Agent Trust Hub on Apr 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: Uses the
infshCLI tool to interact with image generation models, including commands for authentication (infsh login), searching applications (infsh app list), and executing generation tasks (infsh app run). - [EXTERNAL_DOWNLOADS]: References external resources for CLI installation and image processing. It fetches images from user-provided URLs for editing tasks and refers to documentation and install scripts hosted on the vendor's official GitHub repository.
- [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by processing external prompt strings and image URIs. 1. Ingestion points: prompt strings and reference_images URIs provided via JSON input. 2. Boundary markers: No explicit delimiters or warnings provided for external content. 3. Capability inventory: Execution of the
infshbinary via the Bash tool. 4. Sanitization: No sanitization or validation of input strings was observed before execution. - [SAFE]: The skill operates within its stated purpose of providing an interface to Alibaba's image generation models. All identified tools, commands, and external links belong to the vendor's verified infrastructure (inference.sh).
Audit Metadata