The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted external sources (Websites, PDFs, Videos) to generate design specifications, prompts, and code (HTML slides), which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Untrusted data enters the context through the extraction modules in references/extractors/from-website.md, references/extractors/from-pdf.md, and references/extractors/from-video.md.
Capability inventory: The skill uses the extracted data to generate HTML/CSS for presentations (references/connectors/html-slides.md), instructions for AI video generation (references/connectors/heygen-video-agent.md), and design tokens for Figma and paper.design.
Boundary markers: The skill utilizes structured YAML templates and specific instructions in the extraction prompts to constrain the output format.
Sanitization: There are no explicit instructions for sanitizing or escaping content extracted from source materials before it is interpolated into final outputs.

visual-style