Convex Agents Fundamentals
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- PROMPT_INJECTION (LOW): The skill defines a generation action that processes untrusted user input, creating a surface for indirect prompt injection. Ingestion points: prompt argument in the generateReply action. Boundary markers: Absent; user input is interpolated directly into the generation call. Capability inventory: limited to thread.generateText (internal reasoning and display). Sanitization: No input validation or filtering is implemented in the provided example.
- EXTERNAL_DOWNLOADS (LOW): The skill references standard packages (@convex-dev/agent and @ai-sdk/openai) from the npm registry. While the convex-dev scope is not on the explicit pre-approved trusted organization list, these are recognized tools for the Convex developer ecosystem.
Audit Metadata