AI Behavior Architect

What problem does it solve?

Provides a disciplined prompt engineering framework to prevent hallucinations, ensure correct tool usage, and enforce invisible self-auditing for sales and support agents in e-commerce environments.

Core Features & Use Cases

• Structured System Prompts: Defines identity, injected context (e.g., catalog_summary), safety constraints, and required output formats for conversational agents.
• Tool Calling Governance: Enforces explicit instructions for when to call product search or other tools instead of guessing, improving factual accuracy.
• RAG & Audit Protocols: Mandates hallucination handling ("say you don't have that information") and integrates the Assist Score auto-audit via report_assistance(type, score, reasoning) every three user messages.
• Use Case: Update a sales agent's system prompt so it always calls search_products for catalog queries, responds in WhatsApp-friendly format, and quietly reports Assist Score for conversion tracking.

Quick Start

Update the system prompt to enforce product search tool calls and enable invisible Assist Score auditing every three user messages.