Pre-invoke trust check for unknown agents
The agent receives a trust decision including a reliability score and recommendation status for a target endpoint, allowing it to safely decide whether to proceed with an API call.
How it works
The agent extracts the target endpoint URL from the user's request or context. It then calls the check_agent_trust tool with this URL. Since this tool requires no API key, it can be used immediately. The agent analyzes the returned trustDecision, reliabilityScore, and reasons to determine if the target is safe. It should check if the endpoint is registered and if the trust decision is 'recommended'. If the score is low or the endpoint is unregistered, the agent should warn the user or refuse to proceed.
Prompt
Paste this into an agent that has the servers below connected. Adjust the names and dates to your own.
Check the trustworthiness of the following agent endpoint before I invoke it: https://api.example.com/v1/agent. Please provide the reliability score and trust decision.MCP servers it needs
Source
The vendor page this playbook is drawn from, and the passage that supports it.
check_agent_trust( {endpointUrl} ) Pre-invocation trust check for an unknown AI agent or MCP server endpoint. No API key or account required.API Reference — AgentTrust