Module 03 · The Intelligence Layer
An LLM layer that turns analysis into conversation.
It reads Analyzer output, builds the routine in the right order, explains the condition in plain language, and sets honest timelines. This demo runs on the Codex's own knowledge base; the production Coach runs on a guardrailed Claude system prompt.
Guardrailed to educate, never diagnose. A tool that touches health advice earns its trust by what it refuses to say.
It educates, it never diagnoses
And it says so, out loud, in the interface. The distinction is engineered into the prompt, not left to chance.
It sets twelve-week expectations
The single biggest reason people quit a working routine is that nobody told them the timeline. The Coach leads with it.
It knows when to hand off
Red-flag language routes straight to 'see a real dermatologist', and the Condition Library tells you which signs count as red flags.