System Prompt Components

The system prompt is composed of 6 components:

Component	Description	Example
system_message	Agent behavior and role	"You are an agent specialist in..."
instructions	What the agent should do	"You MUST respond to the user..."
expected_output	Format of the response	"Your answer must be concise..."
examples	Input/output examples	Examples of reasoning and outputs
system_extra_message	Additional system context	Extra instructions or constraints
include_date	Include current date	Adds "Weekday, Month DD, YYYY"

All components are assembled using a system prompt template, that can be customized via templates={"system": "..."}. By default, the template concatenates all defined components in a structured format using XML tags.

Example

# pip install msgflux[openai]
import msgflux as mf
import msgflux.nn as nn

# mf.set_envs(OPENAI_API_KEY="...")

class BusinessAgent(nn.Agent):
    model = mf.Model.chat_completion("openai/gpt-4.1-mini")
    system_message = """
    You are a business development assistant,
    focused on helping sales teams.
    """
    instructions = """
    When given a company description, identify potential needs,
    suggest an outreach strategy, and provide a value proposition.
    """
    expected_output = """
    Respond in three bullet points:
    - Identified Needs
    - Outreach Strategy
    - Value Proposition
    """
    system_extra_message = """
    Ensure recommendations align with ethical sales practices.
    """
    config = {"include_date": True, "verbose": True}

agent = BusinessAgent()
print(agent.get_system_prompt())

Expected Output:

<developer_note>
<system_message>

    You are a business development assistant,
    focused on helping sales teams.

</system_message>

<instructions>

    When given a company description, identify potential needs,
    suggest an outreach strategy, and provide a value proposition.

</instructions>

<expected_output>

    Respond in three bullet points:
    - Identified Needs
    - Outreach Strategy
    - Value Proposition

</expected_output>

    Ensure recommendations align with ethical sales practices.


The current date is: Friday, February 20, 2026

</developer_note>

Examples

In-Context Learning (ICL) is a technique where language models learn to perform tasks by observing examples provided directly in the prompt, without any parameter updates. This allows models to generalize from just a few demonstrations.

Few-Shot Learning refers to providing a small number of input-output examples that guide the model's behavior. These examples act as implicit instructions, showing the model the expected format, reasoning style, and output structure.

Benefits of using examples:

Format consistency: The model mimics the structure of your examples
Implicit instructions: Complex behaviors can be demonstrated rather than explained
Reasoning patterns: Show chain-of-thought reasoning for the model to follow
Domain adaptation: Tailor responses to your specific use case

There are three ways to pass examples to an Agent:

Few-shot Example Formats

String ExamplesExample ClassDict Examples

Simple text format:

# pip install msgflux[openai]
import msgflux as mf
import msgflux.nn as nn

# mf.set_envs(OPENAI_API_KEY="...")

examples = """
Input: A startup offering AI tools for logistics.
Output:
- Needs: Supply chain optimization
- Strategy: Highlight cost savings
- Value: Reduce delays with predictive analytics

Input: An e-commerce platform for handmade crafts.
Output:
- Needs: Market visibility
- Strategy: Cross-promotion with eco marketplaces
- Value: Global audience access for artisans
"""

class SalesAgent(nn.Agent):
    model = mf.Model.chat_completion("openai/gpt-4.1-mini")
    system_message = "You are a business development assistant."
    instructions = "Identify needs and suggest strategies."
    expected_output = "Three bullet points: Needs, Strategy, Value"
    examples = examples

agent = SalesAgent()
print(agent.get_system_prompt())

Expected Output:

<developer_note>
<system_message>
You are a business development assistant.
</system_message>

<instructions>
Identify needs and suggest strategies.
</instructions>

<expected_output>
Three bullet points: Needs, Strategy, Value
</expected_output>

<examples>

Input: A startup offering AI tools for logistics.
Output:
- Needs: Supply chain optimization
- Strategy: Highlight cost savings
- Value: Reduce delays with predictive analytics

Input: An e-commerce platform for handmade crafts.
Output:
- Needs: Market visibility
- Strategy: Cross-promotion with eco marketplaces
- Value: Global audience access for artisans

</examples>

</developer_note>

Structured examples with metadata:

# pip install msgflux[openai]
import msgflux as mf
import msgflux.nn as nn

# mf.set_envs(OPENAI_API_KEY="...")

examples = [
    mf.Example(
        inputs="A fintech offering digital wallets.",
        labels={
            "Needs": "Payment integration and trust",
            "Strategy": "Position as secure and easy-to-use",
            "Value": "Simplify digital payments"
        },
        reasoning="Small retailers need trust and ease.",
        title="Fintech Lead",
        topic="Sales"
    ),
    mf.Example(
        inputs="An e-commerce for handmade crafts.",
        labels={
            "Needs": "Visibility and market reach",
            "Strategy": "Partner with eco marketplaces",
            "Value": "Global audience for artisans"
        },
        reasoning="Handmade crafts need visibility to scale."
    )
]

class SalesAgent(nn.Agent):
    model = mf.Model.chat_completion("openai/gpt-4.1-mini")
    examples = examples

agent = SalesAgent()
print(agent.get_system_prompt())

Expected Output:

<developer_note>

<examples>
<example id=1 title="Fintech Lead" topic="Sales">
<input>A fintech offering digital wallets.</input>
<reasoning>Small retailers need trust and ease.</reasoning>
<output>{"Needs":"Payment integration and trust","Strategy":"Position as secure and easy-to-use","Value":"Simplify digital payments"}</output>
</example>

<example id=2>
<input>An e-commerce for handmade crafts.</input>
<reasoning>Handmade crafts need visibility to scale.</reasoning>
<output>{"Needs":"Visibility and market reach","Strategy":"Partner with eco marketplaces","Value":"Global audience for artisans"}</output>
</example>
</examples>

</developer_note>

Dict-based examples are converted to Example objects:

# pip install msgflux[openai]
import msgflux as mf
import msgflux.nn as nn

# mf.set_envs(OPENAI_API_KEY="...")

examples = [
    {
        "inputs": "A startup offering AI tools for logistics.",
        "labels": {
            "Needs": "Supply chain optimization",
            "Strategy": "Highlight cost savings",
            "Value": "Reduce delays with predictive analytics"
        }
    },
    {
        "inputs": "An e-commerce for handmade crafts.",
        "labels": {
            "Needs": "Market visibility",
            "Strategy": "Cross-promotion with eco marketplaces",
            "Value": "Global audience access"
        }
    }
]

class SalesAgent(nn.Agent):
    model = mf.Model.chat_completion("openai/gpt-4.1-mini")
    examples = examples

agent = SalesAgent()
print(agent.get_system_prompt())

Expected Output:

<developer_note>

<examples>
<example id=1>
<input>A startup offering AI tools for logistics.</input>
<output>{"Needs":"Supply chain optimization","Strategy":"Highlight cost savings","Value":"Reduce delays with predictive analytics"}</output>
</example>

<example id=2>
<input>An e-commerce for handmade crafts.</input>
<output>{"Needs":"Market visibility","Strategy":"Cross-promotion with eco marketplaces","Value":"Global audience access"}</output>
</example>
</examples>

</developer_note>