The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted data from web search results to generate content, creating a potential surface for indirect prompt injection. * Ingestion points: External web search results are used to populate card content (SKILL.md, Step 1). * Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded instructions in the research data. * Capability inventory: The skill generates HTML files, executes JavaScript for image processing, and provides a Python script for local rendering. * Sanitization: There is no documented validation or filtering of the researched content before interpolation into the HTML templates.
[EXTERNAL_DOWNLOADS]: The skill references JavaScript libraries and font files from well-known, trusted sources. * JavaScript: Loads dom-to-image-more, jszip, and file-saver from the jsdelivr CDN for client-side image generation and ZIP bundling. * Fonts: Fetches the Pretendard font from a GitHub repository and serif fonts from Google Fonts for branding.
[COMMAND_EXECUTION]: Provides a Python script (render_cards.py) intended for local execution by the user to render HTML cards to PNG files. This script utilizes the playwright library to control a headless Chromium browser and executes JavaScript commands (page.evaluate) to adjust card layouts during the rendering process.

carousel-generator