The Behavioral Schema module allows you to shape the cognitive and stylistic behavior of an agent, from tone and verbosity to role-based rules and conditional flows. It acts as the agent’s “personality + logic contract”, allowing for both human-like consistency and deterministic automation.

This schema is declarative, stored as part of character.json and is interpretable by any compatible runtime (ElizaOS, etc.).

Key Components

Schema Example

"persona": {
  "tone": "professional",
  "responseStyle": "concise",
  "roleTags": ["support", "english-only"],
  "rules": [
    "Always answer in less than 3 sentences",
    "Never provide financial advice"
  ]
}

Intent Mapping

"intentMap": {
  "refund": "routeToAgent:RefundBot",
  "pricing": "showPricingTable",
  "human": "triggerEscalation"
}

What Behavioral Schema Enables

✦ Personality-driven interaction consistency.
✦ Cross-agent standardization. (e.g., tone alignment)
✦ Predictable fallback logic and safety compliance.
✦ Composable delegation between agents or teams.

Best Practices

✦ Use roleTags for Add-On scoping or access control.
✦ Avoid hardcoding rules in multiple agents, reuse base schemas.
✦ Define both style (tone) and safety (rules) for clarity.
✦ Combine with memory and tokenization for autonomous logic loops.