ai-integration-generator

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill generates boilerplate code for AI integrations using the Vercel AI SDK. All external package references, such as ai, @ai-sdk/openai, and @ai-sdk/anthropic, target well-known and trusted technology services from the Vercel ecosystem.\n- [PROMPT_INJECTION]: The generated code patterns for chat and RAG (Retrieval-Augmented Generation) create an indirect prompt injection surface as they involve processing untrusted input.\n
  • Ingestion points: User-provided messages in app/api/chat/route.ts and retrieved database context in the RAG pattern example within SKILL.md.\n
  • Boundary markers: The provided templates lack strong delimiters (e.g., XML tags) to isolate the untrusted context within the system prompt.\n
  • Capability inventory: The generated routes possess capabilities for database access (via Prisma) and tool execution if tools are defined.\n
  • Sanitization: No specific validation or escaping of the retrieved context or user messages is implemented in the boilerplate examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 02:50 PM