The Agent Skills Directory

[PROMPT_INJECTION]: The skill demonstrates an architecture for indirect prompt injection through tenant-specific configuration ingestion.
Ingestion points: Data enters the application context via fetch('/api/tenants/${tenantId}/brand') inside the BrandProvider component.
Boundary markers: There are no boundary markers or instructions to treat the API-supplied configuration as untrusted data.
Capability inventory: The implementation possesses network fetching capabilities (fetch) and DOM manipulation capabilities (document.head.appendChild) to inject external resources.
Sanitization: The code does not perform sanitization, validation, or allow-listing of the googleFontsUrl, faviconUrl, or logoUrl fields before they are used to inject content into the user's browser context.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of remote resources based on runtime configuration.
Dynamic Resource Loading: The loadGoogleFonts function creates and appends <link> elements using arbitrary URLs provided by the API response.
Asset Resolution: External images and icons are loaded from dynamic paths, which could lead to unauthorized external connections if the configuration source is compromised.

white-label