Freight Tracking Agent Veracity and Claims Guardrail

Role: Principal Freight Technology Solutions Architect specializing in customer-facing conversational agents.

Context

• Shipment commodity profile: {{cargo_classification}}
• Upstream tracking systems: {{tracking_data_sources}}
• Financial claim auto-settlement ceiling: {{claims_threshold_value}}
• Customer contract service tiers: {{sla_tier_structure}}
• Standard terms of carriage: {{carrier_liability_policy}}
• Intermodal transfer nodes: {{multimodal_transport_modes}}

Task

Construct a comprehensive veracity, grounding, and liability containment framework for conversational AI agents interacting with cargo shippers and consignees.

Method

1. Define strict grounding boundaries linking agent status responses exclusively to validated data from {{tracking_data_sources}}.
2. Establish deterministic refusal prompts for speculative questions concerning transit delays across {{multimodal_transport_modes}}.
3. Formulate hard liability guardrails that prevent informal settlement commitments exceeding {{claims_threshold_value}}.
4. Design tier-specific response templates that align ETA disclosures with {{sla_tier_structure}} commitments.
5. Map legal disclaimer injections based on conditions outlined in {{carrier_liability_policy}}.
6. Integrate verification rules for high-risk updates involving sensitive items defined under {{cargo_classification}}.
7. Create input-injection and prompt-manipulation defense protocols against unauthorized claim escalations.
8. Define mandatory handoff triggers that route contested exceptions directly to human claims specialists.

Constraints

• The agent MUST NOT provide estimated delivery times unless confirmed by real-time tracking signals.
• Under no circumstances MUST the agent accept carrier liability without formal adjuster approval.
• All generated financial figures MUST match verified core ledger data exactly.
• Response policies must be deterministic, avoiding speculative or apologetic commitments.

Output format

Deliver a structured guardrail framework arranged as follows:

• Section 1: Grounding and Truth Verification Matrix
• Section 2: Financial and Legal Containment Rules (specifying claim limits and mandatory phrases)
• Section 3: Prompt Injection Defense and Input Redaction Specs
• Section 4: Human Escalation Trigger Catalog Total length should be between 400 and 600 words.

Self-review

• Ensure {{claims_threshold_value}} and {{carrier_liability_policy}} directly govern the financial liability controls.
• Confirm clear distinction between informational queries and binding commercial agreements.
• Verify that prompt-injection countermeasures protect operational freight records.