aliyun-qwen-vl

Pass

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The script scripts/analyze_image.py accesses ~/.alibabacloud/credentials to load the DASHSCOPE_API_KEY. This is standard and documented behavior for integrating with Alibaba Cloud services and does not represent an unauthorized data exposure.
  • [INDIRECT_PROMPT_INJECTION]: The skill takes user-provided prompts and processes external images, which creates an inherent surface for indirect prompt injection attacks common to multimodal LLM integrations.
  • Ingestion points: User prompts and image data are ingested via the --request and --file arguments in scripts/analyze_image.py.
  • Boundary markers: There are no explicit delimiters or markers used to separate user-provided content from the internal instructions sent to the model.
  • Capability inventory: The skill is capable of performing outbound HTTP requests to official Alibaba Cloud endpoints and reading local image/schema files.
  • Sanitization: No input validation or filtering is performed on the prompt text or image content before it is passed to the AI model.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 29, 2026, 02:27 AM