The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Cross-Site Scripting) through its data ingestion process.
Ingestion points: The scripts/generate-presentation.js script processes JSON or YAML files to populate HTML templates.
Boundary markers: Absent. There are no instructions or delimiters to prevent the rendering engine from interpreting embedded scripts or HTML tags within the content fields.
Capability inventory: The skill can write files to the local system (fs.writeFileSync) and execute browser-based tasks via Playwright (scripts/export-slides.js), which could be leveraged to execute malicious scripts if the generated HTML is opened.
Sanitization: Incomplete. While escapeHtml is defined, it is selectively applied. Several slide renderers (including repo, image, stats, grid, quote, and comparison) interpolate properties like url, src, title, and value directly into the HTML string without escaping, allowing for attribute breakout and script injection.
[EXTERNAL_DOWNLOADS]: The skill utilizes external dependencies and assets from well-known services.
Dependencies: Requires the playwright package for exporting slides to PNG and PDF formats.
CDNs: The templates/base.html file fetches the anime.js library from Cloudflare's CDN and loads typography from Google Fonts.
[COMMAND_EXECUTION]: The skill provides scripts for local execution.
Node.js Scripts: Includes generate-presentation.js, export-slides.js, and md-to-slides.js which perform file system operations and spawn browser instances for rendering.

presentation-generator