integrate-fusion-agent
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for creating an agent server that exposes application data and actions to an LLM. This creates a surface for indirect prompt injection where data processed by the agent could influence its behavior.
- Ingestion points: Application state is ingested through resources defined in
agentResources.tsusing thecreateAgentResourcefunction. - Boundary markers: The guidance suggests providing clear
descriptionstrings for resources and actions to guide the agent's behavior. - Capability inventory: The server exposes actions defined in
agentActions.ts(e.g.,update_item_status,get_item_details) which can modify application data via thehandlerfunctions. - Sanitization: The skill recommends using Zod schemas for parameter validation in actions and explicitly advises that mutating actions require prior user approval.
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install
@cognite/app-sdkusing a package manager. This is a vendor-owned resource used for legitimate integration with the host application environment.
Audit Metadata