{"version":"2.1.288","anchor":"auto-mode-classifier-overflow-now-triggers-a-compact-and-a-r","canonical_anchor":"auto-mode-classifier-overflow-now-triggers-a-compact-and-a-r","heading":"Auto mode can shorten a long conversation instead of repeatedly blocking actions","tier":"notice","area":"Auto Mode","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.288\/e\/auto-mode-classifier-overflow-now-triggers-a-compact-and-a-r","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.288","markdown":"### Auto mode can shorten a long conversation instead of repeatedly blocking actions\n\nWhen a conversation is too long for auto mode's safety check, Claude Code can now compact it and tell Claude to retry the action afterwards\n\n**Unclear.** What decides whether Claude Code asks for compaction, rather than only blocking, is not settled.\n\n**What**\n\nIn auto mode, a classifier (a separate model check) reviews each action Claude wants to take. When the conversation grows too long for that check to read, the action is blocked.\n\nBefore, the block message said the same action would keep hitting the same limit until the conversation got shorter. Now, in some cases, Claude Code asks for the conversation to be compacted (summarised so it takes up less room). Claude is told the action did not run, that Claude Code will try to compact before the next request, and to issue the action again afterwards if it is still needed. Claude Code also records each time it asks for compaction this way.\n\n**Why**\n\nLong auto-mode sessions get a way to recover instead of having every action blocked until you shorten the conversation yourself.\n\n- Area: Auto Mode\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 2\/5\n- Scope: individual\n- Heads-up: no"}