design-evaluation
Pass
Audited by Gen Agent Trust Hub on Apr 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates entirely through structured natural language instructions and data tables. It does not utilize any shell commands, scripts, or external tool invocations that could pose a security risk. No patterns of prompt injection, data exfiltration, or obfuscation were detected.
- [EXTERNAL_DOWNLOADS]: The skill references external resources from trusted organizations including Cloudflare, Vercel, and Neon. These references are used solely for documentation purposes and do not involve remote code execution or unauthorized data transfer.
- [DATA_EXFILTRATION]: No network operations or sensitive file access patterns were found. The skill processes project-specific data provided by the user within the agent's safe environment.
- [PROMPT_INJECTION]: The instructions are clearly scoped to urban design evaluation. There are no attempts to override system safety filters, extract system prompts, or disregard previous instructions.
- [NO_CODE]: The skill consists of Markdown files containing evaluation frameworks and benchmarks without any embedded executable scripts or configuration files that modify the execution environment.
Audit Metadata