gpt-image-2-image-edit
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection because it ingests external data (images and prompts) that are subsequently processed by an AI model.
- Ingestion points: User-supplied file paths via the
--imageargument and text via the--promptargument inscripts/imagegen.py. - Boundary markers: The skill does not employ boundary markers, delimiters, or system instructions to distinguish between task instructions and data content, nor does it warn the model to ignore instructions embedded within the images.
- Capability inventory: The skill has the ability to perform network requests (GET, POST, PUT) to the
ai-hive.iclip.cnAPI and object storage, and it can write files to the~/Downloads/AiHivedirectory on the local file system. - Sanitization: No validation or sanitization is performed on the content of the provided images or the prompt strings before they are transmitted to the external API.
Audit Metadata