image-generation
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The script
scripts/generate_image.jstransmits a user-provided prompt and an API key todashscope.aliyuncs.com. This is the official endpoint for Alibaba Cloud's DashScope service, and the skill follows standard security practices by sourcing the credential from theDASHSCOPE_API_KEYenvironment variable.\n- [EXTERNAL_DOWNLOADS]: The skill uses Node.js built-inhttpandhttpsmodules to fetch image data from URLs returned by the generation API. It saves these files to the local directory using timestamp-based filenames to prevent collisions.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes output from an external AI model to determine the source URL for image downloads, representing a potential injection surface.\n - Ingestion points: The script
scripts/generate_image.jsparses the JSON response body from the DashScope API to extract image locations.\n - Boundary markers: No explicit boundary markers or instruction-filtering logic is applied to the content returned by the remote model.\n
- Capability inventory: The skill has the ability to perform network GET/POST requests and write files to the filesystem via the
fsmodule.\n - Sanitization: The script validates the target using the
URLconstructor but does not implement a domain-specific allowlist for the downloaded content.
Audit Metadata