cover-hero
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The script
assets/code/src/render-cover.jsusesexecSyncto runnpx -y --no-install hyperframes browser path. This command is used to dynamically locate the Chrome browser executable on the host system to facilitate the Puppeteer rendering process. - [EXTERNAL_DOWNLOADS]: The skill depends on the
puppeteer-coreNode.js package, as defined inassets/code/package.json. The rendering scriptrender-cover.jsalso attempts to locate this dependency within the user's local npm cache (~/.npm/_npx) if not found in the standardnode_modulesdirectory. - [DATA_EXPOSURE]: The skill performs local file system operations, such as reading Markdown files and frontmatter content via
pick-template.jsand writing generated HTML/PNG files into a task-specific directory (imgs/). These actions are required for the skill's core functionality of transforming text content into visual assets. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by reading content from Markdown files (
--articleflag) and user input in interactive mode. This content (titles, keywords, card descriptions) is injected directly into HTML templates. - Ingestion points: Frontmatter parsing in
pick-template.jsfrom user-provided Markdown files. - Boundary markers: None identified; the skill assumes the provided Markdown files are safe.
- Capability inventory: File system writes (
fs.writeFileSync), command execution viaexecSync, and full HTML rendering via Puppeteer. - Sanitization: The
injectContentfunction inpick-template.jsallows raw HTML tags (e.g.,<em>,<br/>,<span>) to be injected into the templates for styling purposes, which represents a potential surface for cross-site scripting (XSS) within the Puppeteer rendering context if the input source is malicious.
Audit Metadata