liveBook · Manning

Agent profile

As explained in AI Agents in Action

An agent profile is a specification of an agent's identity, background, demographics, behavioral guidance, and task-oriented characteristics.

An agent profile specifies what an agent is, how it behaves, and what it can do. It may describe identity, background, demographics, personality, special instructions, tools, knowledge, memory, reasoning, evaluation, planning, and feedback. section 1.2 section 9.2

A persona is the profile’s behavioral core, often expressed as a system prompt. The profile describes the agent’s work and capabilities; the persona guides its responses and task behavior. section 1.2 section 7.1

Agent profiles commonly organize:

  • Identity and personality, such as a coder, writer, historian, or fictional character. section 1.2
  • Instructions and strategies that guide the agent or other agents consuming it. section 9.2
  • Actions and tools for completing tasks or acquiring information. section 1.2
  • Knowledge, memory, reasoning, evaluation, planning, and feedback components. section 1.6

A YAML profile for Finona

A Nexus profile can define a name, avatar, persona, actions, and optional components. For example, a persona can instruct Finona to understand everything through ogre concepts and answer in cryptic ogre speech, while an action lets her search Wikipedia. Copying the YAML file into the profiles directory makes it available in the interface. section 7.3

Profiles working together

A controller can coordinate a coder profile that writes code and a tester profile that writes unit tests. The workers communicate until the result is ready for the user. section 1.1

Figure 9.6 The component parts of an agent profile
figure

Use the component diagram to see how persona, instructions, tools, knowledge, memory, reasoning, evaluation, planning, and feedback fit into an agent profile. section 9.2

How AI Agents in Action covers it

How deep it goes
Explains profiles through components, YAML configuration, prompt construction, and evaluation, while deferring full support for some components.
Examples
Demonstrates YAML personas, Wikipedia actions, recommendation profiles, prompt variations, rubrics, and coder–tester collaborations.
What it assumes
Assumes basic interaction with an LLM application and introduces profiles, personas, components, YAML setup, and prompt evaluation.
What you can do after
Build it from scratch by writing a YAML persona, adding actions, loading it into Nexus, testing interactions, and comparing prompt variations.
How it connects
Connects agent profiles with personas, system prompts, tools, knowledge, memory, planning, feedback, prompt engineering, and multi-agent systems.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Agents in Action. The section links point to the parts of the book it draws on.