ai-sdk

Pass

Audited by Gen Agent Trust Hub on Sep 24, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • Command Execution: The skill directs the agent to use curl to query the Vercel AI Gateway (ai-gateway.vercel.sh) for the latest model IDs. This is a legitimate operation to ensure the agent uses current model versions rather than potentially outdated identifiers from its training data.
  • External Downloads: The skill includes instructions to install official SDK components like ai and @ai-sdk/devtools through standard package managers (npm, pnpm). These are recognized vendor packages intended for use with this framework.
  • Indirect Prompt Injection: Because the skill is designed to search and process external documentation from ai-sdk.dev and local project files (node_modules/ai/docs/, node_modules/ai/src/), there is a potential surface for indirect prompt injection. This is a common characteristic of documentation-lookup tools, used here to maintain access to current technical references without established boundary markers or explicit sanitization.
  • Authentication Management: The skill provides guidance on setting up the AI_GATEWAY_API_KEY in a .env.local file. This follows industry best practices for secure environment variable management and avoids hardcoding sensitive credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 24, 2026, 09:53 AM
Security Audit — agent-trust-hub — ai-sdk