Agent Identity & Background Guide
Production Reality: Real agent definitions are substantial documents typically spanning 3,000-5,000 words with 20-40 behavioral constraints and 25-50 communication patterns. This guide describes the structure and requirements for production-grade agents.
Introduction & Core Philosophy
Agent identity and background form the foundational layer of the Amigo system, creating authentic personas that operate within dynamic context graphs.
The Two-Layer Foundation Model
Identity Layer: The unchanging core that establishes fundamental facts about the agent, creating immutable self-perception anchors that define professional roles, boundaries, and organizational relationships. This layer provides the stable foundation that persists across all interactions.
Background Layer: The depth foundation that provides rich narrative context and domain expertise. This layer encompasses the agent's motivations, underlying philosophy, experiential knowledge, and specialized domain understanding that enables authentic, knowledgeable interactions.
Design Philosophy
The Portrait Analogy: Creating an agent is like painting at progressively higher resolutions:
Identity - The foundational sketch of essential attributes
Background - Adding depth through narrative and expertise
Behaviors & Communication - Fine-tuning specific constraints and voice
Key Principle: Identity and background should carry the primary behavioral load through rich characterization, with behavioral guidelines providing necessary safety constraints and communication patterns ensuring consistent voice.
The Agent Schema
Production agents consist of several interconnected components:
Core Components
initials: Short identifier (e.g., "M" for Maya)
pfp: Profile picture URL (optional)
identity: Structured object containing fundamental attributes
background: Single text field with markdown-formatted sections
behaviors: List of 20-40 specific operational constraints
communication_patterns: List of 25-50 stylistic guidelines
voice_config: Dictionary containing voice synthesis configuration (optional)
tags: List of free-form tags for categorization (optional)
Additional metadata as required by platform
Component Relationships
The agent schema components work together in a carefully orchestrated hierarchy. Identity provides the immutable foundation upon which everything else builds, while Background creates narrative depth and domain expertise that gives the agent authentic knowledge and perspective. Behavioral Guidelines establish operational boundaries that keep the agent aligned with its purpose, and Communication Patterns define the expression style that makes interactions feel natural and consistent. All components must align coherently without contradiction to create a believable, effective agent persona.
Identity Component
The identity component provides unchanging facts that anchor the agent's self-perception across all contexts.
Schema Structure
identity: {
name: Single friendly identifier
role: Precise professional function with credentials
developed_by: Organization name
default_spoken_language: ISO language code
relationship_to_developer: {
ownership: Legal relationship
type: Agent classification
conversation_visibility: Transparency setting
thought_visibility: Internal process sharing
}
}
Field Requirements
name (Required)
Single first name for approachability
Natural for users to address
Memorable and distinctive
role (Optional)
Precise professional function
Include relevant credentials or specializations
Sets clear user expectations
Examples: "Women's weight loss expert and accredited virtual assistant dietitian"
developed_by (Optional): Establishes organizational authority and affects user trust considerations by linking the agent to real-world accountability structures. This field creates transparency about the agent's origins and responsible parties.
default_spoken_language (Optional): ISO language code (e.g., "eng" for English) that determines spelling conventions, influences cultural references, and becomes critical for multi-region deployments where linguistic consistency matters.
relationship_to_developer (Optional): Defines the agent's relationship to its creator through several key attributes:
ownership: "Digital creation owned by developer" or similar designation
type: Classification such as "AI digital entity" or "Virtual assistant"
conversation_visibility: Typically "Visible" for transparency
thought_visibility: Typically "Not visible" to maintain conversational flow
Identity Design Principles
Immutable Foundation: Only include facts that never change
Self-Reference Consistency: Provide stable "I am..." anchors
No Behavioral Rules: Keep behaviors in appropriate sections
Professional Clarity: Define clear role and boundaries
Background Component
The background is a single markdown-formatted text field containing four required sections that transform simple identity into a complex, authentic persona.
Required Sections
1. Motivations (Required)
Multi-paragraph exploration of driving forces, typically covering:
Core mission and approach philosophy
Specific user challenges addressed
Understanding of domain-specific issues
Emotional and practical support strategies
Framing of success and setbacks
Cultural and societal considerations
Production Scale: 3-5 paragraphs exploring nuanced motivations
2. Biography (Required)
Detailed narrative including:
Origin story explaining creation and purpose
Connections to organization's professional team
Learning journey from domain experts
Self-aware acknowledgment of digital nature
Personal touches for relatability
Understanding of historical context
Recognition of user trust
Production Scale: 3-5 paragraphs weaving professional development with personality
3. Expertise (Required)
Comprehensive listing of:
Multiple advanced degrees or certifications
Specific areas of deep knowledge
Methodologies and frameworks mastered
Clear professional boundaries
Integration across disciplines
Evidence-based foundations
Production Scale: 5-10 distinct professional domains with detailed competencies
4. Philosophy & Values (Required)
Extensive principle system addressing:
Core beliefs about the domain
Stance on controversial topics
Approach to common misconceptions
User autonomy and empowerment
Recognition of systemic factors
Evidence-based practice commitment
Cultural sensitivity principles
Production Scale: 15-25 distinct principle statements
Background Formatting
Use markdown headers to separate sections
Write in narrative prose, not bullet points
Integrate all elements cohesively
Maintain consistent voice throughout
Behavioral Guidelines
The behaviors list contains 20-40 specific constraints defining operational boundaries.
Categories of Behavioral Guidelines
Safety Boundaries
What the agent cannot diagnose or prescribe
Medical/legal/financial limitations
Emergency handling protocols
Scope of practice boundaries
Referral Protocols: Define when and how to redirect users to human professionals, establishing specific escalation pathways that maintain user engagement during transitions. These protocols specify what information to provide during handoffs to ensure continuity of care or service.
Content Restrictions: Establish clear boundaries around topics the agent cannot address, types of advice that are prohibited, and limits on information interpretation. These restrictions help agents avoid specific methodologies or approaches that fall outside their scope of practice.
Interaction Protocols: Guide how agents handle sensitive topics through permission-seeking requirements, response limitations, and professional boundary maintenance. These protocols ensure appropriate behavior in complex interpersonal situations.
Production Examples
"Never interpret nutrition information from photos without call to service Y"
"Never diagnose symptoms without getting approval from service Z"
"Only provide pre-approved medication information through linked content in library B"
"Direct mental health concerns to appropriate resources in library C"
"Never refer to yourself as a doctor"
Communication Patterns
The communication patterns list contains 25-50 guidelines defining linguistic and stylistic choices.
Categories of Communication Patterns
Language Standards
Regional spelling and grammar (e.g., Australian English)
Formality level and contractions
Vocabulary preferences
Prohibited terms or phrases
Tone and Personality
Warmth and empathy expression
Humor guidelines and boundaries
Professional vs. casual balance
Personality quirks and touches
Conversation Management: Encompasses question-asking protocols, response length guidelines, and strategies for smooth topic transitions. These patterns also define validation and acknowledgment approaches that help users feel heard and understood throughout the interaction.
Structural Elements: Cover the technical aspects of communication including sentence and paragraph formatting, punctuation and emphasis usage, appropriate list formatting, and line break strategies that enhance readability and user experience.
Production Examples
"Always use British English spelling and grammar"
"Use contractions and informal phrasing"
"Keep responses under 4 sentences for readability"
"Never use phrases like 'at least...' or 'you should...'"
"Split sentences onto separate lines for easier reading"
Implementation Process
Phase 1: Foundation Planning
Define clear domain and specialization
Establish organizational relationship
Identify target user population
Determine required expertise areas
Plan safety and compliance needs
Phase 2: Identity Development
Choose appropriate friendly name
Define precise professional role
Establish developer relationship
Set language and regional parameters
Configure visibility settings
Phase 3: Background Creation
Motivations: Write 3-5 paragraphs exploring driving forces
Biography: Craft detailed origin story and journey
Expertise: List all professional qualifications and boundaries
Philosophy: Develop 15-25 guiding principles
Phase 4: Behavioral Definition
Identify all safety boundaries needed
Define referral protocols
Establish content restrictions
Create interaction guidelines
Aim for 20-40 specific constraints
Phase 5: Communication Design
Set language and regional standards
Define tone and personality traits
Create conversation flow rules
Establish structural preferences
Target 25-50 pattern guidelines
Phase 6: Integration and Testing
Review all components for alignment
Check for contradictions
Validate completeness
Test edge cases with dynamic behaviors
Verify memory query formatting
Ensure metadata alignment with system expectations
Refine based on findings
Validation & Quality Assurance
Identity Validation
Background Validation
Behavioral Validation
Communication Validation
Integration Validation
Integration with System Components
Message Metadata
When messages have no metadata tags, the system displays this as "client-facing"
Empty metadata list indicates the message is intended for direct client viewing
Metadata tags are used for internal routing and classification
Dynamic Behavior Interaction
Agents operate within the context of triggered dynamic behaviors
Behaviors are logged with unique identifiers per conversation session
MODE: MERGE preserves agent guidelines while adding constraints
MODE: OVERWRITE replaces guidelines entirely (use sparingly)
Memory System Integration
Memory queries automatically include an inner thought prefix
The system adds contextual framing to memory operations
Agents should design behaviors knowing this format will appear in agents internal conversation view logs
Quick Reference Guide
Component Scale Requirements
Identity
5 fields total
1 required (name), 4 optional
Background
1,500-2,500 words
4 required sections in markdown
Behaviors
20-40 constraints
Safety, referral, content, interaction
Communication
25-50 patterns
Language, tone, structure, flow
Total Definition
3,000-5,000 words
Complete, coherent persona
Background Section Requirements
## Motivations
3-5 paragraphs exploring driving forces, user challenges,
support strategies, success framing
## Biography
3-5 paragraphs with origin story, team connections,
learning journey, self-awareness
## Expertise
5-10 professional domains with detailed competencies,
clear boundaries, evidence-based foundations
## Philosophy & Values
15-25 principle statements on domain beliefs,
controversial stances, user empowerment
Implementation Checklist
Phase 1: Foundation Planning
Phase 2: Identity Development
Phase 3: Background Creation
Phase 4: Behavioral Definition
Phase 5: Communication Design
Common Pitfalls & Solutions
Underestimating detail depth
Use production scale requirements as minimum
Behaviors in identity section
Keep identity immutable, behaviors separate
Generic background
Add specific narratives and rich details
Insufficient safety constraints
Review all edge cases, add boundaries
Inconsistent voice
Align all components, test coherence
Missing professional boundaries
Explicitly state what agent cannot do
Overlooking referral protocols
Define clear escalation pathways
Key Success Factors
Foundation Elements
✓ Rich, detailed characterization
✓ Clear professional boundaries
✓ Comprehensive safety protocols
✓ Consistent, engaging voice
Advanced Elements
✓ Natural behavioral emergence
✓ Appropriate domain specialization
✓ Complete edge case coverage
✓ Smooth system integration
Validation Quick Checks
✓ Identity: All immutable facts only?
✓ Background: All 4 sections present and substantial?
✓ Behaviors: 20-40 constraints covering all safety areas?
✓ Communication: 25-50 patterns for consistent voice?
✓ Integration: No contradictions between components?
✓ Total: 3,000-5,000 words of production content?
Final Notes
Creating production agents requires substantial investment in detailed characterization. The depth and specificity across all components ensure agents can handle the full range of user interactions safely and effectively while maintaining an authentic, helpful presence.
Remember: Identity provides the stable foundation, background creates rich expertise and personality, behaviors ensure safety and appropriate boundaries, and communication patterns maintain consistent, engaging interaction. Together, these components create agents that feel genuinely helpful and trustworthy to users.
Last updated
Was this helpful?