The Agent Skills Directory

[OBFUSCATION]: The file demo/index.html contains a large Base64-encoded string (encodedHtml) representing the entire HTML structure and styling for the demonstration page. This technique obscures the template's content and logic from static analysis tools.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from external domains, including a Supabase storage bucket for visual assets and reputable CDNs such as cdnjs (Cloudflare), Tailwind, and Iconify for essential runtime scripts and styles.
[DYNAMIC_EXECUTION]: The demo environment dynamically generates page content by decoding Base64 data into an iframe using srcdoc. It also utilizes URL.createObjectURL to convert binary asset blobs into usable URLs at runtime within the sandboxed environment.
[INDIRECT_PROMPT_INJECTION]: The text reveal component in SKILL.md reads textContent from elements and injects it back as HTML structure. While it includes a custom escapeHTML function to sanitize input, it represents an attack surface for processing untrusted content. 1. Ingestion points: element.textContent in SKILL.md. 2. Boundary markers: None. 3. Capability inventory: DOM manipulation via innerHTML. 4. Sanitization: Implements escapeHTML to sanitize text segments before injection.

masked-reveal