The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents the use of components such as Paragraph and Html which accept raw HTML strings via the html prop to render formatted content. An agent following these instructions might take untrusted data from a user or external source and place it directly into these props without sanitization, creating a vector for cross-site scripting (XSS) or other injection attacks in the final rendered output.
Ingestion points: User instructions and external data sources processed by the agent to generate template content (referenced in SKILL.md and references/component-reference.md).
Boundary markers: The skill does not provide specific instructions for delimiting or escaping user-provided content when used within these HTML props.
Capability inventory: The generated HTML may be rendered in email clients, web browsers, or exported as PDF, potentially affecting anyone viewing the document.
Sanitization: There are no explicit instructions or examples for sanitizing user-provided strings before they are assigned to the html prop.

unlayer-elements