reflect-session

To extract reusable learnings from a completed agent session, review the original goal, approach, and outcomes, then convert insights into structured types like Mental Model, Hard Rule, and Gotcha for future recall to prevent repeated mistakes.

The best way to retain categorized session insights is to prompt the user to approve extracted learnings, then retain each approved memory individually using the Hindsight retain mechanism for durable future recall across agent workflows.

Yes, you can categorize decision-making and process improvement insights by structuring them into defined types such as Decision, Technique, Discovery, Hard Rule, and Gotcha during your session review.

Capturing session reflections requires structuring insights by type and presenting them for user approval before retention, ensuring you selectively keep durable learnings rather than automatically storing the entire session history.

You should run a session review to extract learnings during wrap-up workflows after completing coding, debugging, planning, or decision-making sessions where you want to establish better guardrails for future recall.