The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves fetching and analyzing data from untrusted external sources, creating an attack surface for indirect prompt injection.
Ingestion points: The agent is instructed to fetch content from website URLs, parse uploaded PDF brand guides, and analyze video keyframes to populate design system fields (as described in references/extractors/from-website.md, references/extractors/from-pdf.md, and references/extractors/from-video.md).
Capability inventory: The skill can perform web fetches, generate HTML/JavaScript content, and interact with external design and video automation tools such as HeyGen and Figma via APIs or MCP tools.
Boundary markers: The prompts used for extraction lack explicit safety instructions or delimiters to prevent the agent from following potentially malicious commands embedded within the fetched external data.
Sanitization: No sanitization or validation process is described for the extracted design parameters before they are interpolated into code templates or subsequent tool calls.
[DYNAMIC_EXECUTION]: The skill generates executable scripts based on design system parameters.
The HTML Slides connector (references/connectors/html-slides.md) provides templates for generating standalone HTML files that include inline JavaScript and CSS logic derived from the visual style data.
[EXTERNAL_DOWNLOADS]: The skill references external resources and templates from established services.
The documentation for the HTML Slides connector suggests utilizing fonts from Google Fonts via standard HTML link tags and references a design framework hosted on GitHub (zarazhangrui/frontend-slides).

visual-style