The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates the extraction of visual styles from untrusted external sources, including websites, PDFs, and videos. This functionality introduces a potential surface for indirect prompt injection, where an attacker could embed malicious instructions in the source material's metadata, CSS comments, or text content. These instructions might be unintentionally processed and incorporated into the resulting style file, potentially influencing the behavior of AI agents that later consume the style_prompt_full field.
Ingestion points: Website URLs in references/extractors/from-website.md, PDF documents in references/extractors/from-pdf.md, and video files in references/extractors/from-video.md.
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands within the source data during extraction.
Capability inventory: The skill uses web fetching and visual analysis tools to parse content.
Sanitization: The instructions focus on extracting structured design data but do not include steps for sanitizing the input against natural language instructions.
[EXTERNAL_DOWNLOADS]: The documentation contains references to external repositories, specifically the zarazhangrui/frontend-slides project on GitHub. This is provided as a reference for users implementing the HTML Slides connector and is not part of an automated execution or installation flow.
[NO_CODE]: This skill consists entirely of Markdown instructions and YAML configuration. It does not include any executable scripts (Python, JavaScript, etc.), which reduces the risk of direct malicious code execution.

visual-style