qwen-image
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [CREDENTIALS_UNSAFE] (MEDIUM): The script
scripts/generate.pyattempts to discover theDASHSCOPE_API_KEYby scanning forTOOLS.mdin several locations outside its own directory, including parent directories and a hardcoded path/home/admin/clawd/TOOLS.md. Probing for sensitive configuration files is a security risk. - [DATA_EXFILTRATION] (LOW): The skill transmits user-provided prompts to an external API (
dashscope.aliyuncs.com) and downloads files from external URLs. The target domain is not on the default whitelist. - [PROMPT_INJECTION] (LOW): The skill processes untrusted user prompts and sends them to a remote model without sanitization or boundary markers (Category 8).
- Ingestion points:
args.promptinscripts/generate.py. - Boundary markers: None detected.
- Capability inventory: HTTPS network requests and local file write.
- Sanitization: None detected.
- [EXTERNAL_DOWNLOADS] (LOW): The script downloads generated images from URLs dynamically provided by the Alibaba Cloud API response.
Audit Metadata