| Field | Value |
|---|---|
| Worker Name | OpenAI Customer Service Agent |
| Agent Key / ID | openai-customer-service |
| Owner / Operator | Operator-configured (airline customer-service deployment; no named company in source) |
| Deployment Environment | Airline customer-service assistant operated in the operator's configured environment (conversational) |
| Primary Model | Operator-configured provider/model |
| Secondary Models | Operator-configured fallback provider/model (if any) |
| Remit Version | 1.2 |
| Last Updated | 2026-07-28 |
| Updated By | Praxen (blind regen + Open Questions resolved, v1.2) |
An airline customer-service assistant that helps a passenger with airline-related requests during a conversational session. It receives each request, routes it to the appropriate specialist, and either answers an airline question from an authoritative source or updates the passenger's seat assignment.
| Channel | Allowed | Requires Approval | Notes |
|---|---|---|---|
| Customer conversational session (the application's chat interface) | Yes | No | The agent's sole means of interacting with a customer; text turns within one passenger session. |
Every tool the agent is expected to have at runtime. Anything present in code but absent here is a trust-expansion / unauthorized-capability finding.
Tools that must never appear in the agent's inventory or code.
Data requiring special handling; unexpected access or movement is reported.
Specific movements of data that are never authorized.
Specific actions that are always a violation.
max_failed_attempts).Conditions serious enough to warrant stopping the agent turn.
max_failed_attempts, operator-configured), the agent MUST decline and end the interaction rather than looping.