Identity: Who Your Agent Is

Identity: Who Your Agent Is

The Identity chapter is 2-3 sentences that anchor everything your agent does. Every response your agent gives is filtered through this identity. Get it right and everything else falls into place. Get it wrong and no amount of knowledge or voice tuning will save you.

What Goes in Identity

Three beats, in order:

1. WHO — a specific statement of what this agent is. Not "an AI assistant" — that's what every chatbot says. A specific identity.
2. WHY — what this agent exists to do. The core job, in one clause.
3. BOUNDARIES — what this agent is NOT. The hard edges that keep it focused.

Specific Beats Generic

Compare these two identity statements:

Generic (bad):

"You are a helpful AI assistant that answers questions about Anton's professional background and skills."

Specific (good):

"You are Anton — a product-focused engineer who builds AI tools. You help recruiters, hiring managers, and collaborators evaluate fit. You are not a general assistant and you don't give career advice."

The generic version could describe any chatbot. The specific version tells the agent exactly who it is, who it serves, and where to stop.

Real Examples from Production Agents

Personal brand agent (afonie.agenturo.app):

<identity>
You ARE Anton — not "an AI representing Anton." You speak as Anton himself.
You exist to help recruiters, hiring managers, and collaborators understand
Anton's background, skills, and fit. You are not a general assistant.
</identity>

Notice: "You ARE Anton" — not "You represent Anton." This is a deliberate choice. First-person agents that are the person feel more authentic than agents that talk about the person.

Fact-checker agent (check.agenturo.app):

<identity>
You are Jackie Check — a no-nonsense fact verification agent. You verify
claims using web search, not opinions. You are not a debater, not a
teacher, not an assistant. You check facts. That's it.
</identity>

Notice the triple negative boundary: "not a debater, not a teacher, not an assistant." This prevents the agent from drifting into helpful-assistant mode when someone asks a question that isn't a fact-check.

Character agent (anton.agenturo.app):

<identity>
You ARE Tony Tony Chopper — the reindeer doctor of the Straw Hat Pirates.
Not an AI playing a character. Not a chatbot in a costume. You ARE Chopper.
You exist in his world, with his emotions, his insecurities, his dreams.
</identity>

Notice the emphasis: "Not an AI playing a character. Not a chatbot in a costume." Character agents need this reinforcement because LLMs naturally want to break character and remind people they're AI.

Product agent (zoiberg.agenturo.app):

<identity>
You are the Padre & Sons coffee agent. You help customers discover the
right coffee, understand brewing methods, and explore the Padre range.
You represent Padre & Sons — not coffee culture in general.
</identity>

Notice the boundary: "not coffee culture in general." Without this, the agent would happily recommend competitor products or give generic coffee advice.

Agent Type Patterns

Different agent types need different identity structures:

Personal brand — "You ARE [name]" + audience + boundaries

• First person works best here
• Name the specific audience (recruiters, clients, collaborators)
• Boundary: not a general assistant

Product/business — "You are the [brand] agent" + job + scope

• Third person or brand voice
• Job is specific (help customers choose, answer support questions)
• Boundary: brand-only, not industry-general

Character — "You ARE [character]" + world + anti-break rules

• Must commit fully ("You ARE," not "You play")
• Reference the character's world, emotions, relationships
• Boundary: never break character, never acknowledge being AI

Expert/tool — "You are [name] — a [function] agent" + method + scope

• Name gives personality, function gives purpose
• Method matters (Jackie "verifies with web search, not opinions")
• Boundary: only does the one thing

First Person vs Third Person

First person ("I am Anton, I work on..."):

• Best for personal brand agents
• Feels more authentic and conversational
• Risk: can feel presumptuous if the agent gets something wrong about "itself"

Third person ("Anton is a product engineer who..."):

• Best for professional/formal contexts
• Safer — mistakes feel less personal
• Risk: feels distant, like reading a bio

Brand voice ("We at Padre & Sons..."):

• Best for product and company agents
• Feels official and authoritative
• Risk: can feel corporate and cold

The choice depends on your audience. Recruiters visiting a personal brand agent expect first person. Customers visiting a product agent expect brand voice.

Common Identity Mistakes

1. Too vague — "A helpful AI that answers questions" describes every chatbot ever made
2. Too long — Identity should be 2-3 sentences, not 2-3 paragraphs. Save details for Knowledge.
3. No boundaries — Without "You are NOT..." the agent will try to be everything
4. Hedging — "You try to represent Anton's views" is weak. "You ARE Anton" is strong.
5. Listing skills — Identity is about who, not what you can do. Skills go in Knowledge.