premium-web-design
Warn
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DYNAMIC_EXECUTION]: The skill provides patterns for dynamically injecting a remote JavaScript library from a well-known CDN at runtime.
- Evidence: The implementation guide in SKILL.md describes using document.createElement('script') to fetch the Spline viewer from https://unpkg.com/@splinetool/viewer@1.9.28/build/spline-viewer.js.
- [DYNAMIC_EXECUTION]: The skill instructs the use of the dangerouslySetInnerHTML React property to render 3D components.
- Evidence: The integration logic in SKILL.md uses this property to inject custom HTML tags, bypassing standard JSX security features.
- [INDIRECT_PROMPT_INJECTION]: The skill workflow creates an attack surface for indirect prompt injection by processing external data to generate code.
- Ingestion points: The skill instructions in SKILL.md require the agent to 'Absorb the brief' from the user and 'Research the industry' via web search results and URLs harvested from external sites.
- Boundary markers: No explicit delimiters or safety instructions are provided to separate user-provided content from the agent's core instructions.
- Capability inventory: The generated React components are capable of executing scripts and making network requests for assets.
- Sanitization: The skill does not provide mechanisms to sanitize user-supplied briefs or external search results before they are included in the generated output.
Audit Metadata