orange-clean-paper-saas
Fail
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: HIGHOBFUSCATIONEXTERNAL_DOWNLOADSINDIRECT_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 styles, layout, and script logic. Using Base64 to store and load executable content is a form of obfuscation that prevents direct inspection of the code by static analysis tools.- [EXTERNAL_DOWNLOADS]: The demo metadata indemo/source.jsonidentifies several runtime dependencies from established and well-known technology services including Cloudflare (GSAP), Tailwind Labs, Google Fonts, and Iconify. These references are standard for web development and align with the skill's purpose of providing design templates.- [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection as it is designed to process user-provided content anchors, brand information, and product briefs to generate UI designs. - Ingestion points: User-supplied product brief and content anchors via the prompt template in
demo/PROMPT.md. - Boundary markers: Absent; there are no specific delimiters or instructions to ignore embedded instructions in the processed data.
- Capability inventory: Dynamic generation of HTML and JavaScript code which is subsequently rendered in a browser environment.
- Sanitization: Absent; the skill does not explicitly sanitize or escape user-provided content before interpolating it into the design output.
Recommendations
- AI detected serious security threats
Audit Metadata