The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface in demo/PROMPT.md where the agent is instructed to ingest and analyze external content.
Ingestion points: The agent is directed to use https://neuform.ai/pages/fluid-protocol-flow-analytics/index.html as a visual and interaction reference.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present around the reference ingestion.
Capability inventory: The skill is scoped to design and HTML/CSS generation; no high-risk capabilities such as arbitrary shell execution or credential access were detected.
Sanitization: No explicit sanitization or filtering logic is provided for the external reference content.
[EXTERNAL_DOWNLOADS]: The demo configuration in demo/source.json specifies several runtime dependencies from well-known and trusted providers.
Trusted Sources: Runtime scripts are fetched from cdnjs.cloudflare.com, cdn.tailwindcss.com, and code.iconify.design.
Resource Types: The skill uses standard web libraries including Tailwind CSS, GSAP, Three.js, and Iconify Icons.
[SAFE]: The file demo/index.html contains a large Base64-encoded string (encodedHtml) representing the demo's HTML document.
Content Analysis: Decoding the string reveals legitimate HTML, CSS, and JavaScript for a UI component demo titled "Fluid Protocol | Flow Analytics".
Purpose: The encoding is used to package the demo code for injection into a sandboxed iframe via srcdoc, which is a common security pattern for previewing AI-generated designs.

company-logos