The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function includes extracting visual styles from untrusted external data sources such as website URLs, PDF brand guides, and video files (as documented in references/extractors/from-website.md, references/extractors/from-pdf.md, and references/extractors/from-video.md). If these source files contain malicious instructions, they could influence the agent's behavior during the extraction or application phases.
Ingestion points: External URLs, PDF files, and Video files processed by the agent.
Boundary markers: The skill lacks explicit instructions to ignore commands or natural language instructions that might be embedded within the source data being analyzed.
Capability inventory: The skill has the capability to generate HTML/CSS (references/connectors/html-slides.md), create prompts for third-party AI video generators (references/connectors/heygen-video-agent.md), and produce configurations for design tools like Figma.
Sanitization: There is no evidence of sanitization or filtering logic to prevent processed data from influencing downstream prompt generation.
[DYNAMIC_EXECUTION]: The skill facilitates the generation of executable HTML and CSS code through the references/connectors/html-slides.md connector. While intended for presentation slides, this involves generating code based on user-supplied or extracted design parameters.
[EXTERNAL_DOWNLOADS]: The references/connectors/html-slides.md file recommends using an external GitHub repository (zarazhangrui/frontend-slides) and fetching assets from Google Fonts as part of the styling workflow.

visual-style