gpt-image-2-ecommerce-main-image
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions in
SKILL.mdspecify the manual installation of therequestsPython library usingpip3 install requests. - [DATA_EXFILTRATION]: The
scripts/imagegen.pyscript reads local image files provided by the user and uploads them to a remote server (ai-hive.iclip.cn). While this is the intended function of the image generation service, it involves transmitting local data to an external domain not on the standard whitelist. - [COMMAND_EXECUTION]: The skill provides and executes a custom Python script (
scripts/imagegen.py) to manage API communication, file uploads, and task polling. - [PROMPT_INJECTION]: The skill accepts arbitrary text prompts and reference images which are interpolated into requests sent to a remote AI model, creating a surface for indirect prompt injection attacks.
- Ingestion points: User prompts and image files are ingested via command-line arguments in
scripts/imagegen.py. - Boundary markers: No delimiters or "ignore embedded instructions" warnings are present in the agent instructions.
- Capability inventory: The skill has the ability to read local files and perform network operations (HTTP POST/GET).
- Sanitization: No sanitization or validation of the input prompt or image content is performed before processing.
Audit Metadata