ai-sdk-core
Pass
Audited by Gen Agent Trust Hub on Apr 29, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill templates facilitate the interpolation of untrusted user data into AI prompts (as seen in
templates/cloudflare-worker-integration.tsandtemplates/nextjs-server-action.ts), creating a surface for indirect prompt injection. - Ingestion points: Untrusted data is ingested via API request bodies and function arguments.
- Boundary markers: Basic templates lack explicit delimiters to separate user input from system instructions.
- Capability inventory: The skill demonstrates the use of tools capable of making network requests (e.g.,
fetchinreferences/tools-and-agents.md). - Sanitization: The skill proactively addresses this risk by including a
sanitizePromptfunction inreferences/production-patterns.mdas a recommended best practice. - [EXTERNAL_DOWNLOADS]: Correctly references official packages and documentation from trusted organizations such as Vercel, OpenAI, Anthropic, Google, and Cloudflare.
- [COMMAND_EXECUTION]: Uses standard development commands for package management (npm, bun) and includes instructions for using the official Vercel migration utility.
Audit Metadata