The Agent Skills Directory

[Data Ingestion and Sanitization]: The skill demonstrates how to ingest dynamic content from URL parameters (searchParams.get('title')). While this is a standard use case for Open Graph images, it introduces a surface for indirect prompt injection if the resulting image is processed by other downstream AI agents. The skill naturally mitigates some risk by rendering the content into an image format rather than raw text.
[Remote Resource Access]: The skill provides examples of fetching font files and images from external URLs (e.g., fetch for fonts and <img> tags for absolute URLs). These are documented using standard development practices for the Vercel Edge runtime and target official or placeholder domains.
[Command Execution]: The installation instructions use standard package managers (npm, pnpm, yarn, bun) to install well-known libraries (satori, @vercel/og). These commands are typical for developer-focused skills and align with the stated purpose.
[Boundary Identification]: When processing external data like URL parameters, the skill uses basic interpolation. A security best practice when implementing these patterns is to ensure that user-supplied strings are validated or truncated to prevent unexpected layout issues or content injection.

satori