{"version":"2.1.281","anchor":"classifier-compaction-tracking-and-per-agent-cleanup","canonical_anchor":"classifier-compaction-tracking-and-per-agent-cleanup","heading":"Classifier compaction tracking and per-agent cleanup","tier":"internal","area":"Auto Mode","url":"https:\/\/changelogs.core-directive.com\/v\/2.1.281\/e\/classifier-compaction-tracking-and-per-agent-cleanup","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.281","markdown":"### Classifier compaction tracking and per-agent cleanup\n\nClaude Code now records when each agent's safety-classifier history is compacted and can clear it per agent\n\n**What**\n\nCompaction means shortening a stored conversation history so it takes less space. New internal helpers record the times when an agent's classifier history is compacted. Both the number of records and the size of the table are capped, and a cached token count (`prevCacheReadTokens`) is reset. Another helper deletes an agent's compaction records, classifier lines and previous state.\n\n**Why**\n\nThis is internal housekeeping so the records do not keep growing. The finding does not show a visible change for users.\n\n- Area: Auto Mode\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 2\/5"}