The Agent Skills Directory

[OBFUSCATION]: The file demo/index.html contains a large Base64-encoded string (encodedHtml) that stores the complete HTML and JavaScript for the demo. This obfuscates the demo's implementation details from static analysis.
[DYNAMIC_EXECUTION]: The script in demo/index.html performs dynamic assembly and execution of content. It decodes the encodedHtml string and injects it into an iframe using the srcdoc attribute. Furthermore, a script within the decoded HTML processes asset data received via postMessage, decoding Base64 strings and creating binary object URLs to dynamically update the page's styling and resources.
[EXTERNAL_DOWNLOADS]: The demo references several external runtime libraries and assets from well-known services. These include Google Fonts (fonts.googleapis.com), GSAP hosted on Cloudflare's CDN (cdnjs.cloudflare.com), Iconify icons (code.iconify.design), and Tailwind CSS (cdn.tailwindcss.com). Product assets are also fetched from Supabase storage (hoirqrkdgbmvpwutwuwj.supabase.co).
[INDIRECT_PROMPT_INJECTION]: The design of the demo loader establishes an attack surface where untrusted or external data is processed and rendered.
Ingestion points: The encodedHtml variable and the message event listener in demo/index.html serve as points where data enters the rendering context.
Boundary markers: The demo execution is contained within an iframe using the sandbox="allow-scripts" attribute and is governed by a restrictive Content Security Policy (CSP).
Capability inventory: The skill utilizes atob for decoding, URL.createObjectURL for asset management, and direct srcdoc injection for rendering.
Sanitization: There is no evidence of sanitization, filtering, or validation performed on the decoded HTML or asset data before they are processed.

css-border-gradient