web-design-engineer
Pass
Audited by Gen Agent Trust Hub on May 6, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references several industry-standard libraries including React, Babel, Chart.js, D3.js, Lucide, and Popmotion. These are fetched from well-known and reputable CDNs such as unpkg.com and jsdelivr.net. The instructions recommend using pinned versions and integrity hashes, which are security best practices for remote dependencies.
- [COMMAND_EXECUTION]: Instructions guide the agent to read existing project code to extract design tokens and visual context. This is a standard and expected behavior for a coding assistant and is constrained to the user's project scope.
- [PROMPT_INJECTION]: The skill uses instructional markers like 'Non-negotiable Hard Rules' and 'IMPORTANT,' but these are applied to technical implementation details (e.g., variable naming in React, avoiding certain CSS methods for environment compatibility) rather than attempting to bypass safety filters or override the agent's core identity.
- [SAFE]: The 'Anti-AI-Style Clichés' and 'Placeholder Philosophy' sections serve as qualitative guardrails to improve output quality and prevent the agent from fabricating data or assets. The indirect prompt injection surface is addressed by a workflow that requires design system declaration and early draft reviews, providing human-in-the-loop checkpoints.
Audit Metadata