Session Retrospective

Start from observed friction, corrections, repeated work, uncertainty, and successful shortcuts. Distinguish a one-off event from a reusable pattern.

Use available evidence first; ask only questions that change the recommendation. Do not infer a recurring failure from one incident or describe a suspected cause as established. A single event can expose a reusable defect in an instruction; explain that distinction rather than requiring repeated harm before fixing it.