The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to malicious instructions or scripts embedded in the JSON/YAML data it processes.\n
Ingestion points: scripts/generate-presentation.js reads presentation content from local JSON or YAML files.\n
Boundary markers: None. There are no delimiters or instructions to ignore embedded code in the processed data.\n
Capability inventory: scripts/export-slides.js uses Playwright (Chromium) to render the generated HTML for PNG, PDF, and video export. This browser context typically has access to local resources via the file:// protocol.\n
Sanitization: Inadequate. While code and terminal slides use a helper to escape HTML, all other slide types (title, content, two-col, stats, grid, image, quote, comparison) interpolate input properties directly into HTML template strings.\n- [DYNAMIC_EXECUTION]: Untrusted data from the input file is incorporated into executable HTML content. Most slide renderers in scripts/generate-presentation.js fail to use escapeHtml, allowing for the injection of <script> tags, event handlers, or malicious <iframe> elements.\n- [DATA_EXFILTRATION]: Because the export-slides.js script opens the generated HTML using the file:// protocol in Playwright, any injected JavaScript can use browser APIs to read sensitive local files and potentially exfiltrate them if the execution environment allows network requests.

presentation-generator