macrodata-memory-maintenance

Distill daily conversations by running an end-of-day routine that extracts relevant facts from sessions, updates memory state files, and prunes stale context to reduce memory bloat. This process preserves morning readiness while keeping memory concise and actionable.

End-of-day memory maintenance is an automated process to distill conversations, update state files like today.md and workspace.md, and prune completed items. You need it when agent memory accumulates stale context that reduces actionability.

Update state files and prune completed items by running a maintenance routine that modifies today.md, workspace.md, and human.md while removing stale information. The workflow preserves morning readiness and logs decisions for an auditable journal.

Yes, memory maintenance operates without external dependencies. The routine independently distills conversations, updates markdown state files, and maintains a safety-oriented maintenance journal with no additional components required.

Keep an auditable maintenance journal by logging decisions and surfacing uncertainties during end-of-day memory upkeep. This safety-oriented workflow ensures all pruning and state file updates are recorded for morning review.