Prompts & patterns

Updated May 19, 20261 min read

Single-tool calls are easy. The interesting agent behaviour happens when several tools chain into a flow — diagnosing, narrating, deciding, acting.

This page collects the canonical patterns. Each is a short recipe with the prompt, the tool sequence, and what the agent typically returns.

Operational

Diagnose a stuck shipment

Shipment stuck in the same status for too long? Pull the shipment, the activity timeline, the carrier-side system logs, then decide whether to reprocess or reassign.

Tools: get_shipmentget_shipment_activitylist_shipment_system_logsget_shipment_system_log → (reprocess_shipment | reassign_shipment | refresh_label)

Reschedule a delivery on customer request

Customer calls in asking to push the delivery date. Look up the shipment and reschedule the delivery slot.

Tools: get_shipmentschedule_delivery

Quote and create from natural language

User describes an order in plain English. Resolve tenant + merchant, get rates, create as a draft, confirm.

Tools: get_accessible_tenantsget_merchantsget_shipping_ratescreate_shipment(draft=true)confirm_shipment

Analytics

Daily ops standup

Pull the operational summary, layer in failures and exceptions, narrate as a one-paragraph briefing the ops lead can read in 30 seconds.

Tools: report_dashboard_summary + report_failed_delivery_analysis + report_exception_history

Why did SLA slip last week?

Start from SLA compliance, drill into carrier performance and failure patterns, propose what to investigate next.

Tools: report_sla_compliance + report_carrier_performance + report_failed_delivery_analysis


More patterns coming

As we see more real-world agent traffic, we'll add patterns for cross-border quoting, returns triage, manifest-time-window decisions, and others. Have a flow you'd like to see documented? Email us via Carriyo.com/contact.