qwencloud-image-generation
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for environment verification, such as checking Python versions and detecting API key types using
python3 -c. It also suggests the use ofnpx skills addfor installing or updating related components from the vendor's repository. - [EXTERNAL_DOWNLOADS]: The skill fetches model configurations and catalogs from
alioth-intl.alicdn.com, an Alibaba Cloud CDN domain. These downloads are used to synchronize model availability and settings with the vendor's remote infrastructure. - [DYNAMIC_EXECUTION]: The
scripts/gossamer.pyscript usessubprocess.runto execute an internal update check script (check_update.py). This is used to notify the agent of available updates within the QwenCloud ecosystem. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and reference images which are passed to remote image generation models.
- Ingestion points: User prompt string and reference image URLs/paths in
scripts/image.py. - Boundary markers: Absent; prompts are directly interpolated into the API request JSON.
- Capability inventory: Subprocess execution in
gossamer.py, network operations to DashScope API and Alibaba CDN inqwencloud_lib.py, and file system writes for image downloads. - Sanitization:
qwencloud_lib.pyincludes asanitize_diagnosticfunction to redact URLs (containing signatures) from error messages, but does not sanitize user prompts sent to the model.
Audit Metadata