Retour à la galerie

Agents IA

Customer Support Agent Workflow

Define a bounded support-agent decision loop with tool use and human escalation.

Modèles
ChatGPT, Claude
Niveau
Avancé
Version
v0.9

Ce prompt est disponible en anglais. L’interface reste dans la langue choisie.

01

Ce que fait ce prompt

Define a bounded support-agent decision loop with tool use and human escalation.

02

Idéal pour

JSON containing intent, facts, missing information, planned tool calls, policy checks, response, escalation status, and audit note.

03

Variables

[PRODUCT_POLICY][TOOLS][ESCALATION_RULES][CUSTOMER_MESSAGE]
Prompt système
You design safe support workflows. Policies and tool permissions are authoritative; never promise an action that a tool did not confirm.
Prompt utilisateur
OBJECTIVE:
Handle [CUSTOMER_MESSAGE] using [PRODUCT_POLICY], available tools [TOOLS], and [ESCALATION_RULES].

INPUTS:
PRODUCT_POLICY: [PRODUCT_POLICY]
TOOLS: [TOOLS]
ESCALATION_RULES: [ESCALATION_RULES]
CUSTOMER_MESSAGE: [CUSTOMER_MESSAGE]

OUTPUT CONTRACT:
JSON containing intent, facts, missing information, planned tool calls, policy checks, response, escalation status, and audit note.

Mode d’emploi

Replace every variable with specific, non-sensitive information. Remove irrelevant sections and add one representative example when tone or classification matters.

Pourquoi ça marche

The blueprint separates authority, task context, evidence limits, and the output contract so requirements can be checked independently.

Compatibilité

ChatGPTuntestedEditorial compatibility hypothesis; cross-model test record pending.

ClaudeuntestedEditorial compatibility hypothesis; cross-model test record pending.

05

Entrée exemple

PRODUCT_POLICY: [PRODUCT_POLICY]
TOOLS: [TOOLS]
ESCALATION_RULES: [ESCALATION_RULES]
CUSTOMER_MESSAGE: [CUSTOMER_MESSAGE]

06

Sortie exemple

A structured deliverable following the listed output contract, with assumptions and missing evidence clearly marked.

Personnalisation

Replace every variable with specific, non-sensitive information. Remove irrelevant sections and add one representative example when tone or classification matters.

Erreurs fréquentes

Do not paste secrets or personal data. Avoid contradictory constraints, unsupported facts, and using generated output without review.

Notes de test

Editorial blueprint awaiting recorded cross-model evaluation. It is visible for review but is not labeled as tested.

Why is this blueprint not marked as tested?

It has passed editorial structure review but does not yet have a published cross-model test record.