design-taste-frontend

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The instructions are focused on guiding the agent's aesthetic and technical output. While they use strong language (e.g., 'non-negotiable', 'hard rules'), these are design constraints and best practices for the task, not attempts to bypass safety filters or override system-level instructions.
  • [EXTERNAL_DOWNLOADS]: The skill provides installation commands for standard, industry-recognized UI libraries such as Fluent UI (Microsoft), Carbon (IBM), Radix UI, and shadcn/ui. These are well-known services and do not pose a security risk. Documentation links point to official domains like apple.com, microsoft.com, and digital.gov.
  • [DATA_EXFILTRATION]: No patterns of sensitive file access (e.g., SSH keys, credentials) or unauthorized data transmission to remote servers were identified.
  • [COMMAND_EXECUTION]: The skill mentions shell commands for package management (e.g., npm install) and component initialization (npx shadcn@latest init), which are routine for the stated purpose of frontend development.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 10:25 PM
Security Audit — agent-trust-hub — design-taste-frontend