solar-duotone-bold

Warn

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTION
Full Analysis
  • [OBFUSCATION]: The file demo/index.html contains a large Base64-encoded string assigned to the variable encodedHtml. This string encodes the entire HTML structure, CSS, and JavaScript logic for the Neuform demo, which prevents direct static analysis of the demo's behavior without decoding.
  • [DYNAMIC_EXECUTION]: The demo loader in demo/index.html dynamically decodes the encodedHtml string and injects the resulting content into an iframe using the srcdoc attribute. This allows for the runtime execution of dynamically assembled code.
  • [DYNAMIC_EXECUTION]: The decoded demo content includes scripts that perform runtime compilation of WebGL shaders using gl.compileShader for the background animations.
  • [INDIRECT_PROMPT_INJECTION]: The skill relies on detailed reference designs and prompts (demo/PROMPT.md) to guide the agent's output. This represents an attack surface where instructions embedded in external design references could influence the agent's code generation logic.
  • Ingestion points: Reference URLs and interaction descriptions in demo/PROMPT.md and demo/source.json.
  • Boundary markers: Absent; the agent is instructed to match the reference's logic and behavior closely.
  • Capability inventory: The skill possesses the capability to generate and execute HTML/JS/WebGL in a browser context.
  • Sanitization: Absent; the agent is directed to treat external references as the primary fidelity target.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 16, 2026, 07:33 AM
Security Audit — agent-trust-hub — solar-duotone-bold