vercel-ai-sdk
Pass
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [Prompt Injection] (LOW): The documentation demonstrates indirect prompt injection surfaces where user-provided input is interpolated directly into LLM prompts without sanitization or delimiters.
- Ingestion points: req.json() in Object-Generation.md and Completion.md.
- Boundary markers: Absent; examples use direct template literals.
- Capability inventory: Uses streamText and generateObject which can perform tool calls and generate structured data.
- Sanitization: Absent in the provided code snippets.
- [EXTERNAL_DOWNLOADS] (LOW): The skill references external dependencies and repositories, including the resumable-stream npm package and the vercel/ai GitHub repository. While Vercel is a trusted organization, these references introduce external logic into the agent's context.
Audit Metadata