eve-agent-builder
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill mentions the use of initialization commands such as
npx eve@latest init .andnpx skills add vercel/eve. These commands fetch and execute code from Vercel, a recognized and trusted service provider. These operations are standard practices for setting up the Eve framework in a development environment. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of agents that ingest and process application data, creating a potential attack surface for indirect prompt injection. This is addressed through a structured security framework outlined in the documentation:
- Ingestion points: Data enters the agent context through tools calling existing application services and UI context metadata passed from the frontend.
- Boundary markers: The production checklist mandates that instructions include identity, behavior, and escalation rules, and specifically directs the developer to redact secrets and personal data from tool outputs.
- Capability inventory: The skill uses a modular "slots" architecture (tools, skills, subagents) where each capability is scoped to the authenticated user and tenant context.
- Sanitization: Input validation is enforced via Zod schemas for all tool definitions, and server-side authorization must be re-checked inside every service called by the agent.
Audit Metadata