aura-asset-images
Warn
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: MEDIUMPROMPT_INJECTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
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 with inline JavaScript, effectively hiding executable code from plain-text security inspections. - [DYNAMIC_EXECUTION]: The skill uses
demo/index.htmlto dynamically assemble and render designs within a sandboxediframeusing thesrcdocattribute andpostMessageAPI. This process involves decoding the Base64 template and injecting local runtime scripts into the frame at execution time. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-controlled data such as brand names, copy, and subject matter to generate responsive HTML documents.
- Ingestion points: User-provided parameters in
demo/PROMPT.mdused for design generation. - Boundary markers: Absent. There are no explicit instructions to the agent to treat embedded user content as data only or to ignore instructions within it.
- Capability inventory: The skill generates and executes full HTML/JS documents.
- Sanitization: Absent. The provided logic does not demonstrate validation or escaping of user-provided strings before they are interpolated into the generated code.
- [REMOTE_CODE_EXECUTION]: The demo environment fetches runtime dependencies including Tailwind CSS, GSAP, and Iconify. As documented in
demo/source.json, these are sourced from well-known services and content delivery networks such as Cloudflare and Iconify.
Audit Metadata