ai-sdk-6
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references documentation and canonical code examples from Vercel's official GitHub repository. Evidence includes documentation links in SKILL.md and references/examples.md directed to github.com/vercel/ai. These resources are managed by a well-known service provider.
- [INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing external user messages which are then passed to an AI model with tool-execution capabilities.
- Ingestion points: The POST handler in SKILL.md ingests user messages via req.json().
- Boundary markers: The instructions explicitly recommend using validateUIMessages to delimit and verify incoming content.
- Capability inventory: The skill enables model-driven tool execution through streamText and ToolLoopAgent.
- Sanitization: The documentation provides specific mitigations including the use of Zod-based inputSchema for tools, server-side authorization checks, and the requirement of version 6.0.202 as a security floor to ensure support for HMAC-signed tool approvals.
Audit Metadata