The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to create HTML documents by interpolating untrusted data into predefined templates, creating a vulnerability to content injection.
Ingestion points: External data is placed into the PAGE_TITLE, SUBTITLE, and main content sections of index.html (SKILL.md).
Boundary markers: There are no boundary markers or delimiters defined to isolate user-provided content from the HTML structure.
Capability inventory: The skill's primary function is to write generated HTML files to the local file system.
Sanitization: The skill does not provide instructions to escape, sanitize, or validate data before it is written into the HTML template, which could allow malicious scripts or HTML to be executed when the file is opened in a browser.
[EXTERNAL_DOWNLOADS]: The HTML template references external resources from well-known and trusted providers.
Evidence: The base template loads the Tailwind CSS library from cdn.tailwindcss.com and typography from fonts.googleapis.com (Google Fonts).

infopage