The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing and using established tools such as Marked.js, Pandoc, gomarkdown, Jekyll, and Hugo. These downloads target official package registries (NPM, RubyGems, Go) and the official GitHub repositories of well-known organizations including markedjs, jgm, gomarkdown, jekyll, and gohugoio.
[COMMAND_EXECUTION]: The skill documents standard CLI workflows for document conversion and static site generation. This includes commands for initializing projects, running local development servers (e.g., hugo server, jekyll serve), and executing build processes using pandoc and marked.
[INDIRECT_PROMPT_INJECTION]: The skill documents the conversion of Markdown (untrusted data) into HTML, which creates an attack surface for XSS or prompt injection. The documentation proactively addresses this by providing a mandatory security evidence chain:
Ingestion points: Markdown files and strings processed via CLI tools and programmatic APIs.
Boundary markers: Not present in basic examples, but the skill emphasizes security boundaries in its warning sections.
Capability inventory: Subprocess calls for CLI tools (Pandoc, Hugo, Jekyll) and script execution for Node.js and Go libraries.
Sanitization: The skill provides explicit guidance on using DOMPurify, sanitize-html, and js-xss for Node.js, bluemonday for Go, and the --sandbox flag for Pandoc to mitigate risks.

markdown-to-html