The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user descriptions and interpolates them directly into HTML/React rendering templates without escaping or sanitization. * Ingestion points: The input argument in the skillRender function (found in skill-render.js) which receives natural language descriptions. * Boundary markers: Absent. User-provided strings are directly inserted into HTML template literals in generate-html.js (e.g., ${content.title}). * Capability inventory: The skill utilizes Puppeteer (headless browser) in generate-html.js and the Remotion bundler (Webpack) in remotion/render.js to process and execute the generated code. * Sanitization: Absent. There is no evidence of HTML entity encoding or script stripping for user-provided strings before they enter the browser context.
[COMMAND_EXECUTION]: The skill utilizes powerful rendering engines that involve code execution. * generate-html.js launches a Puppeteer browser instance to render and screenshot HTML content. * remotion/render.js invokes a bundling process that runs Webpack to compile React code at runtime. * These operations are standard for infographic generation but represent a significant capability tier processing data derived from user input.

infographic-generator