Proactive Memory

Turn repeated failures, corrections, and recurring requests into durable follow-up work.

• When an agent hits the same failure twice, create a short memory note with the trigger, failed approach, corrected approach, and verification result.
• Convert important corrections into board tasks or scheduled follow-ups instead of leaving them only in chat history.
• After a user changes a preference, record the new preference in memory and mention the update in the next relevant run.
• Pair with scheduler for check-ins such as "retest this flow tomorrow" or "review this integration every Monday".
• Pair with hierarchy goals when the lesson belongs to a team, project, or organization rather than one agent.
• Keep entries concise: problem, correction, evidence, next trigger.
• Prefer updating an existing memory cluster over creating duplicate notes for the same issue.
• If a correction changes a standard operating pattern, recommend enabling or updating the matching skill pack.