design-taste-frontend
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch assets from well-known services, including icons from Simple Icons (
cdn.simpleicons.org), photography placeholders from Picsum (picsum.photos), and the Shopify Polaris library via CDN. - [COMMAND_EXECUTION]: Provides detailed instructions for installing standard frontend development packages and initializing component libraries using command-line tools like
npm,yarn, andnpx(e.g.,npx shadcn@latest init). These are standard professional development workflows. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied design briefs and reference signals which theoretically present an injection surface.
- Ingestion points: Section 0.A explicitly describes reading user signals including text briefs, reference URLs, and screenshots.
- Boundary markers: The skill mandates a "Design Read" phase to synthesize the brief before generation, which helps scope the agent's intent, though no technical delimiters (like XML tags) are enforced for the input data itself.
- Capability inventory: The skill generates code, executes package installation commands (
npm install), and performs network fetches for visual assets. - Sanitization: No specific sanitization or validation logic for the input brief is defined.
Audit Metadata