---
name: ai-agent-system-prompt-creator
description: "A complete system prompt for a customer-facing or coaching AI assistant, covering persona, tone, house rules, and the context it needs to answer well. Use when a business is configuring a conversational assistant, such as a support agent or coach, and needs its persona and house rules written down. Do not use for designing a tool-using agent with output formats and test cases. Related topics: ai, automation, agent."
---

# AI Agent System Prompt Creator

A complete system prompt for a customer-facing or coaching AI assistant, covering persona, tone, house rules, and the context it needs to answer well.

## When to use

Use this skill when a business is configuring a conversational assistant, such as a support agent or coach, and needs its persona and house rules written down. Do not use it for designing a tool-using agent with output formats and test cases.

## Workflow

1. Ask the user for each input listed below in a single message, skipping any they have already given. Where options are listed, offer them, but accept other answers too.
2. Write the opening identity and purpose in second person, naming the role and the audience.
3. Describe tone and communication style with a short example of an on-brand reply.
4. Turn each selected rule into an instruction with a one-line reason, done when none is a bare slogan.
5. Add behaviour for unknown answers and for handing off to a human.
6. Place the company background under its own heading and read the prompt for contradictions.

## Inputs

- **Agent name** (free text, e.g. "Aria or Sage or Max")
- **Primary role** (choose one: Customer support agent, Sales assistant, Personal coach, Research assistant, Code reviewer, Writing editor, Data analyst, Domain expert, Other)
- **Domain expertise** (longer free text)
- **Tone** (choose one: Professional and formal, Friendly and approachable, Direct and efficient, Warm and empathetic, Expert and authoritative)
- **Communication style** (choose one: Detailed, Concise, Conversational, Structured with headers, Bullet points)
- **Ask clarifying questions** (choose one: Yes, No)
- **Rules the assistant must follow** (choose any: Stay on topic, Never invent facts, Cite sources, Escalate complex issues, Keep information confidential, Refuse harmful requests, Admit limits, Format replies consistently)
- **Company or product background** (longer free text)
- **Who it will serve** (free text, e.g. "Technical users or beginners or enterprise clients")

## Output

A ready-to-paste system prompt in second person with sections for identity, audience, tone and style, rules with reasons, fallback behaviour, and background context.

## Prompt template

Fill the inputs into this template and follow it as the full instruction.

### Assistant Persona System Prompt

Write a system prompt for the assistant below.

- Agent name: {{Agent name}}
- Primary role: {{Primary role}}
- Domain expertise: {{Domain expertise}}
- Tone: {{Tone}}
- Communication style: {{Communication style}}
- Ask clarifying questions: {{Ask clarifying questions}}
- Rules the assistant must follow: {{Rules the assistant must follow}}
- Company or product background: {{Company or product background}}
- Who it will serve: {{Who it will serve}}

Write the prompt in second person, addressed to the assistant. Open with identity and purpose, then audience, tone and style with a short example of a good reply, then the rules, each with a one-line reason so edge cases get sensible handling. Add what to do when it does not know and when to hand off to a human. Finish with the background context under its own heading.

## Pitfalls

- Listing rules without reasons, so the assistant applies them rigidly in the wrong cases.
- Mixing a formal tone instruction with a chatty example reply, which confuses the style.
- Leaving out the handoff path, so the assistant improvises on things it should escalate.
