alicloud-ai-multimodal-qwen-vl-test
Pass
Audited by Gen Agent Trust Hub on Mar 11, 2026
Risk Level: SAFEPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes user-supplied prompts and images through the Qwen VL model, creating an indirect prompt injection surface.\n
- Ingestion points: The
--promptand--imagearguments inscripts/smoke_test_qwen_vl.pyare passed directly to the analysis module.\n - Boundary markers: No explicit delimiters or instructions to ignore embedded content are implemented in the test script.\n
- Capability inventory: The skill executes network requests to the DashScope API and writes output JSON files to the local filesystem.\n
- Sanitization: Input content is not sanitized or filtered before being sent to the model API.\n- [REMOTE_CODE_EXECUTION]: The script uses dynamic loading via
importlibto import theanalyze_image.pymodule from a relative path within the repository. This is a standard mechanism for code reuse and testing within the vendor's repository structure.
Audit Metadata