The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to interpret external user briefs and reference signals (URLs, screenshots, brand assets), which introduces a potential surface for indirect prompt injection.
Ingestion points: Section 0.A and 11.B define inputs from user briefs, reference URLs, and existing site audits.
Capability inventory: The agent is instructed to generate React/Next.js code, install packages, and configure project styling.
Boundary markers: The skill lacks explicit instructions to treat external brief content as untrusted or to ignore embedded instructions within those briefs.
Sanitization: No specific input sanitization or validation protocols are defined.
[EXTERNAL_DOWNLOADS]: The skill references and suggests downloading resources from well-known and trusted services.
Fetches documentation and configurations from official GitHub repositories of trusted organizations such as Microsoft (Fluent UI), Google (Material Web), IBM (Carbon), Shopify (Polaris), and GitHub (Primer).
Recommends using cdn.simpleicons.org for SVG icons and picsum.photos for placeholder images.
References a CDN-hosted script from Shopify's official domain for Polaris integration.
[SAFE]: The skill adheres to development best practices, recommending official packages and secure configuration patterns, such as self-hosting fonts and isolating interactive logic in client-side components.

design-taste-frontend