model-routing
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a framework for classifying user requests and interpolating them into model-specific prompt structures, establishing an attack surface for indirect prompt injection.
- Ingestion points: User request classification in SKILL.md.
- Boundary markers: Includes a 'Hard Boundary' section requiring the agent to prioritize live model capabilities and authorized roles over researched preferences.
- Capability inventory: No code execution, file system access, or network tool invocation capabilities are defined in the skill.
- Sanitization: Employs structured prompt templates and explicit instructions to preserve specific user-defined identity and layout traits while routing to specialist models.
- [EXTERNAL_DOWNLOADS]: The skill references documentation and research benchmark URLs from several external AI service providers to support its model routing recommendations. Mentions include resources from OpenAI, Google, Midjourney, ByteDance, xAI, MiniMax, Fal.ai, and Topaz Labs. All referenced domains are well-known technology or AI research services.
Audit Metadata