The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests markdown data from files or standard input and converts it to HTML using the markdown-it library with the html: true configuration. This creates a vulnerability where malicious instructions or scripts embedded in a document (e.g., via HTML tags) are rendered and executed in the browser context.\n
Ingestion points: Content is read from user-provided files or stdin in scripts/md-render and src/render.mjs.\n
Boundary markers: None. The skill does not implement delimiters or warnings to ignore instructions embedded within the document content.\n
Capability inventory: The script scripts/md-render writes rendered HTML to a temporary file and uses the open or xdg-open commands to launch the system's web browser.\n
Sanitization: Missing. The rendering logic in src/render.mjs explicitly enables HTML rendering and does not apply filtering or escaping to the input markdown.\n- [COMMAND_EXECUTION]: The Bash script scripts/md-render executes the open command on macOS or xdg-open on Linux systems to automatically display the generated HTML file in a web browser.\n- [EXTERNAL_DOWNLOADS]: The skill downloads the Mermaid.js library from a well-known CDN at runtime to support diagram rendering. It also executes npm install within its script to ensure its required Node.js dependencies are available.

md-render