company-logos
Warn
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: MEDIUMOBFUSCATIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The demo implementation in
demo/index.htmlencapsulates an entire HTML document and its associated scripts within a large Base64-encoded string (encodedHtml). This technique obscures the actual page structure and behavior from static analysis. - [DYNAMIC_EXECUTION]: The skill performs runtime decoding of the
encodedHtmlvariable indemo/index.htmland dynamically populates thesrcdocattribute of an iframe. This allows the execution of arbitrary scripts and HTML content that is not visible in the plain text of the file. - [EXTERNAL_DOWNLOADS]: The demo references and loads resources from established external services, including typography from Google Fonts (
fonts.googleapis.com,fonts.gstatic.com) and Fontshare (api.fontshare.com). It also utilizes established libraries for animation and icons through Cloudflare's CDN (GSAP, Three.js) and the official Iconify domain. - [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external assets and dynamically generated content.
- Ingestion points: The skill dynamically fetches asset files and processes complex HTML payloads through
demo/index.html. - Boundary markers: The demo is restricted within an iframe using
sandbox="allow-scripts"to isolate its execution. - Capability inventory: The implementation uses
fetchto load resources andpostMessageto communicate data between the host environment and the sandboxed demo. - Sanitization: Content is isolated within a sandboxed
srcdocenvironment, which helps mitigate potential risks from the processed data.
Audit Metadata