The Agent Skills Directory

[PROMPT_INJECTION]: The skill documents a pattern for rendering structured data using dangerouslySetInnerHTML, which presents an indirect prompt injection surface if the input data is sourced from untrusted users.
Ingestion points: Data objects passed to schema generation functions in SKILL.md, such as article.title and product.name.
Boundary markers: The examples use JSON.stringify() for serialization, which provides character escaping but no semantic boundaries or warnings to ignore embedded instructions.
Capability inventory: The skill is limited to metadata generation and DOM injection; it does not contain capabilities for arbitrary command execution, file system access, or network exfiltration.
Sanitization: There is no evidence of explicit sanitization or validation of the input strings before they are rendered into the JSON-LD script block.
[SAFE]: The skill uses official and well-known libraries including next, @vercel/og, and schema-dts for its core functionality.
[SAFE]: All external URL references are to well-known and reputable services such as Google Fonts and schema.org.

seo-meta