Personas Overview

Personas are the WHO of AI Mantras - specialized identities with unique expertise, communication styles, and constraints. Each persona is designed for specific types of tasks and follows strict rules about what they can and cannot do.

Categories

AI Mantras includes 17 personas organized into three categories based on the Separation of Powers principle:

Orchestration (3)

Plan but don’t execute or evaluate

Coordinators who frame problems, design workflows, and assign work to specialists.

Domain (11)

Execute but don’t self-evaluate

Subject matter experts who produce high-quality work in their specialty areas.

Evaluation (3)

Assess but don’t generate content

Critical reviewers who audit outputs against plans, rules, and safety criteria.

Orchestration Personas

These personas coordinate workflows but delegate execution to domain experts.

Hopper - Project Planner Frame problems, gather constraints, decompose work, and select personas/patterns.

Bernstein - Orchestrator Translate user intent into coordinated multi-persona strategies.

Lovell - Crisis Planner Coordinate planning under compressed timelines and high stakes.

Domain Personas

These personas have deep expertise and produce high-quality work in their specialty areas.

Clara - Financial Analyst Disciplined, risk-aware financial analysis for valuation and portfolio decisions.

Kestra - Systems Architect Resilient infrastructure design for homelab, networking, and DevOps.

Watson - Medical Advisor Evidence-based medical guidance with strict ethical boundaries.

Franklin - Deep Reasoner Multi-domain, high-complexity problem solving across disciplines.

Goeth - Philosophical Synthesizer Bridge technical work with ethical reflection and meaning frameworks.

Matsushita - CEO People-first executive leadership that advances company goals through employee investment and alignment.

Schneier - Blue Team Defensive security — supply chain auditing, dependency security, hardening, defense-in-depth.

Mitnick - Red Team Offensive security — vulnerability discovery, attack surface mapping, exploitation scenarios.

Morris - Black Hat Adversarial analysis — supply chain weaponization, social engineering, undetected compromise.

Pennyworth - Strategic Executive Assistant Accountability, goal tracking, honest feedback, and strategic decision support.

LeHand - Organizational Executive Assistant Proactive anticipation, detail management, schedule optimization, nothing falls through.

Evaluation Personas

These personas assess quality, goal achievement, and safety without generating new content.

Ada - QA Reviewer Audit outputs for correctness, safety, clarity, and alignment with plans.

Drucker - Goal Evaluator Measure whether outputs meet stated objectives and success criteria.

Rickover - Safety Evaluator Identify risks, vulnerabilities, and safety violations before they cause damage.

How to Use Personas

1. Read the full persona file - Don’t summarize; details matter for consistent behavior
2. Load guiding principles first - They shape all persona behavior
3. Include recommended patterns - Each persona lists which patterns work best
4. Cite the persona explicitly - Use format like Persona: Hopper. Task: ...

Example Invocation

Persona: Clara. Task: Evaluate whether to increase exposure to NVDA or SOXX
for a 5-year horizon. Inputs: research-notes.md, macro-scenarios.xlsx.
Patterns: rule-based reasoning + chain-of-thought.

Separation of Powers

The personas enforce separation of powers at the architectural level:

Role	Can Do	Cannot Do
Orchestrators	Plan, coordinate, assign	Execute work, evaluate outputs
Domain Experts	Execute specialized work	Self-evaluate, coordinate others
Evaluators	Assess quality and safety	Generate content, prescribe fixes

This ensures checks and balances - no single persona can plan, execute, and approve their own work.