Weekly Reflect

To prune stale entries from agent memory, run a review of recent activity logs and articles to consolidate patterns, then move outdated notes into specific topic files and reduce the main memory index to a short summary.

Consolidating agent memory from weekly run logs involves scanning the last 7 days of activity to extract patterns and outcomes, then merging those insights into refreshed topic notes and recording actions in a daily log file.

Memory log analysis for maintenance works by reading recent run logs and skill health metrics to identify declining quality trends, extracting outcomes, and reorganizing detailed notes into topic-specific files for quick navigation.

Yes, you can focus knowledge pruning on a specific topic area by setting the area-of-focus variable, which restricts the memory consolidation and log review processes to that targeted domain instead of the entire agent memory.

You need existing memory directories like memory/MEMORY.md, memory/logs/, and articles/ available for scanning. The review reads these paths to extract recent activity and update memory/topics/ with refreshed notes.

You should avoid a weekly review if your agent lacks at least 7 days of accumulated run logs or articles, as the process depends on scanning recent activity to identify patterns and prune outdated information.