banner-design
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data such as headlines, CTA text, and brand assets to generate HTML/CSS banner designs. While this introduces an indirect prompt injection surface, the skill focuses on layout generation and image capture, which limits the potential for instruction override.\n
- Ingestion points: User-supplied content and brand parameters collected in Step 1 of
SKILL.md.\n - Boundary markers: None explicitly defined in the design instructions for interpolating user input.\n
- Capability inventory: Uses browser research, image generation, and screenshot tools to finalize assets.\n
- Sanitization: No specific escaping or sanitization of user-provided strings is mentioned before they are rendered in the banner templates.
Audit Metadata