generating-images
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions guide the agent to gather design context by reading external project files, including Markdown, MDX, JSX, and CSS files. This process creates a surface for indirect prompt injection where malicious content within those files could influence the agent's image generation prompts or its actions when modifying project code to wire up the images.
- Ingestion points: Instructions in
SKILL.mdfor analyzing sibling images, surface content (MDX/Markdown), and brand tokens (CSS/config) to inform the image generation prompt. - Boundary markers: The instructions do not specify any delimiters or safety warnings to ensure the agent ignores instructions that may be embedded in the context-gathering data.
- Capability inventory: The skill uses a Python script (
scripts/generate_image.py) executed via the shell and performs file system writes and modifications across the project workspace. - Sanitization: There is no documentation of sanitization or filtering for the content pulled from external project files before it is used in prompt construction.
Audit Metadata