The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs a shell script (scripts/convert-screenshot-to-avif.sh) that uses ffmpeg and ffprobe to process image assets. It also utilizes the Cloudflare wrangler CLI to manage R2 object storage uploads and verification.
[EXTERNAL_DOWNLOADS]: The generated documentation templates fetch established frontend libraries, including daisyUI, Tailwind CSS, Mermaid, and Alpine.js, from well-known services like the jsdelivr CDN.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources like GitHub PR and issue descriptions.
Ingestion points: External content is interpolated into the [data-content-root] section of the output HTML document.
Capability inventory: The skill has the capability to execute local image processing tools (ffmpeg) and deployment tools (wrangler).
Sanitization: The accompanying HTML template (assets/universal-single-file-template.html) uses .textContent for rendering user-supplied strings such as comments and quotes, which prevents Cross-Site Scripting (XSS) in the generated documents.
Boundary markers: The workflow follows defined visual patterns and logical sections to separate instructions from data.
[CREDENTIALS_UNSAFE]: The documentation includes explicit security guidance regarding secret management, specifically instructing users not to paste Google OAuth secrets or Cloudflare tokens into the HTML or chat, and to use secure provider forms instead.

kf-g-html-document-universal-single-file