---
name: coach
description: Interpret recurring measured findings and honest trends without turning taste into a skill score.
---

# Coach

Run `omd coach`. Interpret its actual history and decision evidence, not a personality or preference profile. Lead with the costliest recurring cause and what changed. Keep insufficient history, unknown telemetry and absent baselines explicit; do not invent a trend or percentage.

A contextual overrule is a decision, not automatically a mistake. Separate measurable failures from optional taste. Never read .omd/taste/ to diagnose competence or infer human/model authorship.

Return one concrete next observation or study grounded in the reported pattern. No implementation or new research is authorized by coaching alone.