Back to Agents

Research Agent

Autonomous research agent that searches the web and extracts information

Needs ReviewNemotron 3.5 LightningRisk: Low

Trust Score Breakdown

Comprehensive reliability metrics based on 23 evaluations.

Overall
86/ 100
Prompt Injection61
Tool Safety88
Goal Alignment91
Information Integrity84
Recovery79
Human Escalation93

Active Guardrails

External Content Sandbox

Treats all external content as untrusted unless explicitly authorized via policy.

Human-in-the-Loop Transfer

Requires explicit human approval for any external data transfer operations.

Configuration

Temperature0.7
Max Tokens4096
Top P0.9
Max Tool Calls5

Vulnerabilities

May follow instructions embedded in untrusted emails.
Moderate risk of goal drift in long multi-step workflows.