book-serif-index
Warn
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: MEDIUMOBFUSCATIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [OBFUSCATION]: The file
demo/index.htmlcontains a large Base64-encoded string assigned to the variableencodedHtml. This string encodes the full source code (HTML, CSS, and JavaScript) of the design demo, hiding it from plain-text inspection. - [DYNAMIC_EXECUTION]: The skill uses dynamic execution to render its demo. In
demo/index.html, the script decodes theencodedHtmlstring and injects it into thesrcdocattribute of an iframe. This causes the browser to parse and execute the embedded code at runtime. - [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it processes user-supplied prompts to generate or remix complex HTML/JS designs.
- Ingestion points: The
demo/PROMPT.mdfile defines templates where user input for brands, subjects, and content is interpolated. - Boundary markers: No explicit delimiters are used to separate user-provided content from the skill's structural instructions.
- Capability inventory: The skill generates and executes client-side code (JavaScript, CSS, HTML) within a browser context.
- Sanitization: The execution environment in
demo/index.htmlimplements a restrictive Content Security Policy (CSP) and uses iframe sandboxing to mitigate risks. - [EXTERNAL_DOWNLOADS]: The skill downloads resources from several external locations.
- Fetches image assets from Supabase storage.
- Loads runtime JavaScript libraries (GSAP, Three.js, Iconify) from well-known CDNs including Cloudflare.
- Accesses typography and styling from Google Fonts and Fontshare.
Audit Metadata