The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The generated HTML output references external assets from fonts.googleapis.com (Google Fonts) and cdn.jsdelivr.net (Prism.js). These are well-known, trusted services commonly used for providing fonts and syntax highlighting libraries.
[DATA_EXFILTRATION]: The html_renderer.py script includes an _embed_logo function that reads local file data from a path defined in the skill's configuration (logo_url) and embeds it as a base64 string in the HTML footer. This allows for the accidental or malicious exposure of sensitive local files, such as configuration secrets or private keys, if the logo_url is set to an unauthorized path.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process long-form markdown documents which may contain untrusted content. While the parser sanitizes HTML to prevent cross-site scripting, the skill does not implement boundary markers or instructions to the AI agent to ignore natural language commands hidden within the processed data.
Ingestion points: Markdown content is ingested through markdown_parser.py from files or standard input.
Boundary markers: Absent. There are no delimiters or specific warnings to the agent to treat the document as data only.
Capability inventory: The skill can read local files and write files to the disk.
Sanitization: Performed for HTML tags, but no logic exists to filter or escape instructions targeting the LLM.

md-document