The Agent Skills Directory

[OBFUSCATION]: The file demo/index.html contains a large Base64-encoded string (encodedHtml) that decodes to the HTML and JavaScript source for the interactive demo. This encoding is used for bundling and packaging the demo content and does not appear to contain malicious logic.
[DYNAMIC_EXECUTION]: The demo runner in demo/index.html dynamically constructs the source for a sandboxed iframe by decoding an internal template and injecting script paths at runtime. This uses the srcdoc attribute and postMessage for asset hydration, which is a common pattern for sandboxed previews.
[INDIRECT_PROMPT_INJECTION]: The design prompts in demo/PROMPT.md encourage the agent to ingest content from external reference URLs, which could theoretically be manipulated by an attacker to influence the generated code.
Ingestion points: The prompt refers to external source URLs (e.g., neuform.ai) as design references in demo/PROMPT.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the interpolation logic.
Capability inventory: The skill is used by the agent to generate interactive HTML/CSS code.
Sanitization: No sanitization of the reference content is performed by the skill itself.

skeuomorphic-ui