The Agent Skills Directory

[OBFUSCATION]: The file demo/index.html contains a large Base64-encoded string (encodedHtml) that stores the full HTML and JavaScript source code for a design demo. This technique obscures the implementation details from static text-based analysis.
[DYNAMIC_EXECUTION]: The script in demo/index.html decodes the encodedHtml string at runtime and injects it into an iframe's srcdoc property. This dynamic assembly and execution of code is a concern as it bypasses simple scanning of the file's visible content.
[EXTERNAL_DOWNLOADS]: The skill configuration in demo/source.json references external resources from established services, including Google Fonts, Tailwind CSS, GSAP (via Cloudflare), and Iconify. These are utilized to provide styling and interaction for the demo.
[INDIRECT_PROMPT_INJECTION]: The prompt instructions in demo/PROMPT.md direct the agent to use a remote URL from a third-party domain as a high-fidelity reference. This introduces a surface where external content could potentially influence the agent's code generation.
Ingestion points: The agent is instructed to read a reference snapshot URL in demo/PROMPT.md.
Boundary markers: None identified.
Capability inventory: The skill generates front-end code (HTML/CSS) but does not have system-level capabilities like file writing or shell execution.
Sanitization: No evidence of sanitization for the external reference content.

skeuomorphic-ui