qwen-image-2
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references external resources for installation and documentation, including installation scripts from GitHub (github.com/inference-sh) and documentation from inference.sh. These are standard resources for the well-known service associated with the skill.
- [COMMAND_EXECUTION]: The skill uses the Bash tool restricted to the 'belt' command namespace. It provides examples for logging in, searching for apps, and running image generation models using the CLI tool.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data in the form of image prompts and external image URLs (reference_images). It utilizes structured JSON inputs for the CLI commands, which helps in delimiting user data from the execution environment.
- Ingestion points: User-provided strings in the
--inputargument ofbelt app run(SKILL.md). - Boundary markers: Commands use JSON formatting to encapsulate user-provided prompts and URLs.
- Capability inventory: Limited to executing the
beltCLI via the Bash tool. - Sanitization: Standard CLI argument parsing by the
belttool is expected.
Audit Metadata