gsap

Warn

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: MEDIUMOBFUSCATIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [OBFUSCATION]: The file demo/index.html contains a large Base64-encoded string (encodedHtml) that stores the complete HTML and JavaScript source for the interactive design demo.
  • [DYNAMIC_EXECUTION]: The skill implements a sandboxed preview mechanism in demo/index.html that decodes the encodedHtml string and executes the resulting code using the srcdoc attribute of an iframe at runtime.
  • [INDIRECT_PROMPT_INJECTION]: The instructions in demo/PROMPT.md direct the agent to use both the included sandboxed demo and an external Neuform showcase URL as primary interaction and visual references. This establishes a surface for external content to influence agent behavior.
  • Ingestion points: demo/PROMPT.md references the external URL https://neuform.ai/pages/aura-fluid-drag-interactions/index.html and the bundled demo/index.html file.
  • Boundary markers: None present in the prompt instructions.
  • Capability inventory: The skill is scoped to web development and GSAP code generation; it does not possess access to sensitive system commands or network-writing tools.
  • Sanitization: No specific sanitization or filtering logic is present for the referenced reference content.
  • [EXTERNAL_DOWNLOADS]: The skill uses assets and runtime dependencies originally sourced from well-known and reputable services, as documented in demo/source.json.
  • Runtime libraries (GSAP, Three.js, Iconify) are bundled from Cloudflare and official Iconify CDNs.
  • Style dependencies are fetched from Google Fonts and Fontshare.
  • Project-specific image assets are retrieved from Supabase storage.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 21, 2026, 01:31 AM
Security Audit — agent-trust-hub — gsap