course-ebook-publishing
Warn
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill uses
node:vmto execute JavaScript code extracted via regex from an HTML file (loadCourseFromIndexHtml). Executing code dynamically from local files carries risk if those files are malicious or compromised.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from awindow.COURSEobject found in HTML files to generate documents.\n - Ingestion points: Data is loaded from
window.COURSEwithin theloadCourseFromIndexHtmlfunction documented inSKILL.md.\n - Boundary markers: The skill does not employ specific delimiters or instructions to ignore embedded commands within the ingested course data during markdown composition.\n
- Capability inventory: The skill uses
spawn('pandoc', ...)for document conversion,chromium.launch()(Playwright) for PDF rendering, and file system write operations.\n - Sanitization: The skill filters out quiz-related items for policy reasons but lacks general sanitization or escaping for markdown or HTML injection in the final documents.\n- [COMMAND_EXECUTION]: The skill executes external CLI tools like
pandocviaspawnand launches a headless browser via Playwright. While standard for the skill's purpose, these are powerful capabilities that could be abused if the input data is compromised.
Audit Metadata