high-contrast-skeuomorphic-clean
Warn
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: MEDIUMOBFUSCATIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The file
demo/index.htmlcontains a large Base64-encoded string assigned to the variableencodedHtml. This string decodes to a complete HTML document including JavaScript, which obscures the logic and behavior of the demo from direct inspection. - [DYNAMIC_EXECUTION]: The demo implementation in
demo/index.htmlutilizessrcdocto dynamically load and execute the contents of theencodedHtmlstring within a sandboxed iframe. Additionally, the JavaScript within the decoded payload employsatob()andURL.createObjectURL()to dynamically generate blobs and handle assets received via thepostMessageAPI. - [INDIRECT_PROMPT_INJECTION]: The skill acts on design requirements and product briefs provided as input, which represents a surface for indirect prompt injection where instructions could be hidden in design descriptions.
- Ingestion points: External design briefs, subject tags, and product concepts processed by the agent when applying the skill instructions defined in
SKILL.mdanddemo/PROMPT.md. - Boundary markers: The skill lacks explicit delimiters or instructions for the agent to ignore embedded commands within the processed design data.
- Capability inventory: The skill is scoped to visual design and layout; it does not contain instructions for network operations, sensitive file access, or system command execution.
- Sanitization: No evidence of input validation or sanitization is present for the data ingested during the design generation process.
Audit Metadata