grapesjs-studio-sdk

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements an AI Chat integration that processes project context and user-supplied messages.
  • Ingestion points: ChatRequestBody containing messages and project context in rules/plugins/ai/ai-backend.md.
  • Boundary markers: Present; the implementation supports system prompt customization with preambles and postambles to define AI behavior in rules/plugins/ai/ai-backend.md.
  • Capability inventory: The AI agent can execute tools (e.g., getWeather, generateImage) and trigger editor commands like core:preview as shown in rules/plugins/ai/ai-backend.md and rules/plugins/ai/ai-chat.md.
  • Sanitization: Implicitly handled by the underlying Vercel AI SDK; no explicit sanitization logic is detailed in the skill scripts.
  • [EXTERNAL_DOWNLOADS]: Fetches configurations, media assets, and fonts from well-known services.
  • Fetches placeholder images from picsum.photos and placehold.co.
  • Loads font files from fonts.gstatic.com in rules/configuration/fonts.md.
  • References widely used libraries (jQuery, Bootstrap, reset-css) from established CDNs including cdnjs.cloudflare.com and cdn.jsdelivr.net.
  • [REMOTE_CODE_EXECUTION]: The skill configuration allows for the dynamic loading of scripts from cdn.jsdelivr.net to support third-party plugin functionality.
  • Example: cdnScript options in rules/plugins/components/fslightbox.md, rules/plugins/components/lightGallery.md, rules/plugins/components/swiper.md, and rules/plugins/rte/tinymce.md point to official versions of FsLightbox, lightGallery, Swiper, and TinyMCE.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 03:30 PM
Security Audit — agent-trust-hub — grapesjs-studio-sdk