ai-persona-engine_skill

Overview

This skill builds emotionally intelligent AI personas for voice and chat roleplay using actor-direction prompts rather than technical checklists. It helps creators craft realistic, resistant characters that behave like real people in short, grounded turns. The engine includes a five-layer prompt architecture, audit tooling, and a learning loop to evolve personas over time.

How this skill works

You define a persona through five layers: a brief universal foundation, an elemental archetype (Fire, Water, Air, Earth), a difficulty modifier, conversation-intelligence rules, and character details. The engine produces two outputs per persona: a compact actor-style voice prompt (for voice models) and a fuller system prompt (for simulation and analysis). After each interaction the skill runs an audit node that scores homeowner lines on brevity, authenticity, consistency, enthusiasm leak, and pacing, storing results for automated prompt patches.

When to use it

• Designing voice-first characters for roleplay or sales simulations.
• Creating short, realistic responses for voice assistants or IVR.
• Testing how a persona resists persuasion with minimal language.
• Iterating persona behavior automatically from call audits.
• Producing both a performance-focused voice prompt and a technical spec.

Best practices

• Favor actor direction over prescriptive rules; motivate behavior, don’t enumerate it.
• Keep the universal foundation to 4–5 concise rules max.
• Enforce the Golden Rule: one thought per turn; cut commas and secondary ideas.
• Tune difficulty (1–5) to control word economy and patience.
• Always include the anti-performance voice note: flat, natural, guarded tone.
• Use the audit results to create prompt patches, not manual rewrites.

Example use cases

• A training rig for reps to practice with resistant homeowner archetypes.
• Voice agents for roleplay apps that require believable short replies.
• Automated persona evolution where audits generate incremental prompt patches.
• Testing product messaging against archetypal objections (smokescreens).
• Rapidly prototyping character-driven NPCs for interactive fiction.