The Agent Skills Directory

[OBFUSCATION]: The skill uses Base64 encoding and gzip compression to embed the beautiful-mermaid library directly into generated HTML files. While documented as an optimization for creating self-contained, single-file documentation, this technique obfuscates the library's source code from direct inspection.
[DYNAMIC_EXECUTION]: The skill utilizes (0, eval)(code) to execute the decompressed content of the beautiful-mermaid library at runtime. This allows the diagram engine to be loaded without external script references or a build step, but represents a dynamic execution pattern.
[COMMAND_EXECUTION]: The agent is instructed to run several local Python and Node.js tools for post-processing and auditing the generated HTML, including html_font.py, frontend-antipattern.py, html-visual-gate.mjs, and html-slop-fix.py. It also uses npx serve to host the resulting files for local preview.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied markdown and wiki content to generate complex HTML structures and interactive elements. This creates a surface for potential injection where malicious input could attempt to influence the layout or include unauthorized scripts in the generated documentation site.

docs-site-macos