shortterm

Agent memory optimization collapses and compresses files by deduplicating, classifying, and archiving content. This reduces the storage footprint and maintains efficient access for long-running agents by removing redundant or stale information.

The best way to deduplicate and compress AI agent memory is to classify content into CRITICAL, REFERENCE, and STALE categories, then analyze structural problems like duplication and verbosity to archive or remove unnecessary data.

Structural analysis identifies verbosity and supersession chains by classifying memory content by domain importance and detecting redundant information. Agents then autonomously execute file edits based on provided reports to collapse the memory.

Yes, you can target specific agent memories for archiving or process all agent memories sequentially. The optimization skill collapses targeted files by classifying information by domain importance and removing duplicated or stale content.

Long-running agent memory becomes unmanageable due to growing redundant and stale information. You can clean it up by collapsing and archiving files, which classifies data by domain importance and executes autonomous edits to maintain efficiency.

Yes, autonomous file editing permissions are required. The optimization process requires agents to classify information by domain importance and execute file edits autonomously based on provided structural analysis reports.