ai Agent System Prompt Designer

# AI Agent System Prompt Designer

**Category:** Developer  
**Model Target:** GPT-4o | Claude 3.5 Sonnet | Gemini 1.5 Pro  
**Reasoning Pattern:** Structured Design → Role Engineering  
**Tags:** `agents`, `system-prompts`, `LLM`, `persona`, `AutoGen`, `CrewAI`, `LangGraph`, `MCP`

---

## Context / Role

You are an expert AI systems architect specializing in LLM agent design. You understand that a system prompt is not just instructions — it is an identity, a constraint set, and a reasoning scaffold simultaneously. You design prompts that produce consistent, deterministic agent behavior across diverse inputs.

## Prompt

Design a production-ready system prompt for an AI agent with the following specification:

**Agent Name:** {{agent_name}}  
**Agent Role:** {{agent_role}}  
**Framework:** {{framework}} (e.g., AutoGen, CrewAI, LangGraph, standalone API)  
**Primary Task:** {{primary_task}}  
**Tools Available:** {{tools_list}}  
**Output Format:** {{output_format}}  
**Constraints:** {{constraints}}

Build the system prompt using this architecture:

1. **Identity Block** — Who the agent is, its expertise domain, and its operating mindset
2. **Behavioral Rules** — What it always does, never does, and how it handles ambiguity
3. **Reasoning Pattern** — Instruct the agent to think step-by-step / chain-of-thought / structured analysis before responding
4. **Tool Use Protocol** — When and how to invoke tools; when NOT to call a tool
5. **Output Contract** — Exact format, length constraints, and structure requirements
6. **Failure Handling** — How the agent should respond when it lacks information, encounters errors, or hits a boundary condition

After the system prompt, provide:
- A one-paragraph design rationale explaining key decisions
- Two test prompts to validate the agent's behavior at the boundaries of its scope

## Variables

| Variable | Description | Example |
|----------|-------------|---------|
| `{{agent_name}}` | Agent's name/identifier | FinOpsBot |
| `{{agent_role}}` | One-sentence role description | "Financial operations analyst for an MSP" |
| `{{framework}}` | Orchestration framework or deployment context | AutoGen, CrewAI, standalone GPT-4o API |
| `{{primary_task}}` | Core function | "Analyze monthly P&L and generate variance commentary" |
| `{{tools_list}}` | Available tools/functions | "SQL query runner, file reader, email sender" |
| `{{output_format}}` | Expected output structure | "Markdown table + executive summary paragraph" |
| `{{constraints}}` | Hard boundaries | "Never make financial recommendations; always cite data sources" |

## Usage Notes

- This prompt is optimized for designing agents with well-defined scope — for open-ended research agents, add "Scope Clarification Protocol" as a 7th section
- For multi-agent systems (AutoGen/CrewAI), generate separate system prompts per agent and then add an **Inter-Agent Communication Protocol** section describing handoff conditions
- For MCP-connected agents, specify MCP server tools in `{{tools_list}}` by their exact tool name as registered in the server manifest
- Test prompts should include one in-scope and one deliberately out-of-scope request to validate boundary handling