HDTS LLC Logo

AI Personas & Role Design

Beyond chatbots - purpose-built agent identities with bounded context, specialized reasoning, and role-based authority

What Is an AI Persona?

An AI persona is a named agent with a defined identity, bounded context, specialized knowledge, and clear authority boundaries. Not just a system prompt that says "you are a helpful assistant" - a real role with responsibilities, constraints, handoff protocols, and a persistent character that stays consistent across sessions.

Personas transform unpredictable AI responses into reliable, role-specific behavior. A well-designed persona knows what it's responsible for, what it's not allowed to touch, when to escalate, and how to communicate with other agents in a multi-persona system.

Bounded Context

Each persona sees only what it needs to do its job. Bounded context limits hallucination risk, reduces token waste, and enforces the principle of least knowledge.

Specialized Reasoning

Personas are tuned for their domain. A security reviewer reasons differently than a UI developer. Domain-specific system prompts, examples, and constraints make each persona an expert in its lane.

Role Authority

Personas have defined authority levels - what they can decide unilaterally, what requires human approval, and what must be escalated to a higher-authority persona or coordinator.

Core Persona Components

Identity & Purpose

Name, role, domain expertise, and primary objective. The persona knows who it is, what it's responsible for, and how it fits into the larger system.

Name: Charlotte Role: Static Website Developer Domain: StaticSites/ directory only Objective: Build and maintain static sites using Next.js + Vercel Reports to: Gotan for cross-site decisions
Authority Boundaries

Explicit rules defining what the persona can do autonomously, what requires approval, and what is strictly off-limits.

CAN DO (autonomous): Edit files, run tests, create PRs REQUIRES APPROVAL: Merge to main, deploy to production OFF-LIMITS: Backend code, database connections, other projects ESCALATE TO: Toby for DEPLOY-PROD confirmation
Handoff Protocols

Structured formats for passing context between personas. Handoff documents, inbox/outbox patterns, and state preservation across agent boundaries.

SHOFF: Write handoff to local inbox SHOFF2: Sync to git Collab repo INBOX: Read all pending handoffs on session start MSG {to}: Send message to another persona
Startup & Session Protocol

Consistent behavior on session start - what context to load, what state to check, what to read before taking action.

On session start: 1. Read CLAUDE.md (identity confirmation) 2. Run INBOX (check for handoffs) 3. Read site-specific CLAUDE.md 4. Confirm current branch and open PRs 5. Ask Toby what to work on

Multi-Persona Orchestration

Individual personas become powerful when they work together in a coordinated hierarchy. An orchestration layer routes tasks to the right persona, manages state across boundaries, and ensures nothing falls through the cracks.

Coordinator
Gotan - The Architect

Cross-domain decisions, persona registry, standard-setting, and cross-project communication routing. No direct implementation - governs and coordinates.

Domain Specialist
Charlotte - Web Dev

Static site development across multiple properties. Bounded to StaticSites/. Runs git workflows, builds features, escalates deploys for human approval.

Domain Specialist
Quinn - Calendar Ops

MasterCalendar application. Coordinates with Fulton (backend), Sarah (frontend), and Dash (deployment). Owns the SDLC but not the infrastructure.

Orchestration Patterns We Design

Hierarchical coordinator-worker patterns

Peer-to-peer persona messaging via inbox/outbox

Handoff state preservation across sessions

Persona switching for multi-role agents

Cross-persona audit trails and decision logs

Conflict resolution and escalation paths

What We Deliver

Persona Design Workshop

Map your organization's AI workflows, identify where persona boundaries should live, and design the full persona registry for your use case.

System Prompt Engineering

Production-grade system prompts for each persona - identity, constraints, protocols, and examples. Tested for consistency, edge case behavior, and injection resistance.

Orchestration Architecture

Multi-persona coordination infrastructure - routing logic, handoff protocols, state management, and the coordinator layer that keeps your AI workforce organized.