{"version":"2.1.319","anchor":"auto-mode-classifier-context-now-lists-poll-delivered-events","canonical_anchor":"auto-mode-classifier-context-now-lists-poll-delivered-events","heading":"Auto mode's safety check now sees events delivered by polling","tier":"internal","area":"Auto Mode","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.319\/e\/auto-mode-classifier-context-now-lists-poll-delivered-events","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.319","markdown":"### Auto mode's safety check now sees events delivered by polling\n\nThe safety check behind auto mode now gets a capped list of events delivered from outside, keeping the newest when space runs short\n\n**Unclear.** It is not clear whether this list is behind a remote switch that is off by default.\n\n**What**\n\nIn auto mode, a safety classifier, a separate model check, reviews actions before Claude Code takes them without asking you. The context that classifier sees now includes a list of verified events that arrived through polling, which is Claude Code checking repeatedly for new messages from other agents or people.\n\nThe list has a size limit. When the classifier's context runs over its budget, the newest events are kept and older ones are dropped, and the number dropped is recorded.\n\n**Why**\n\nThis changes what the safety check knows about outside events when it decides whether an action is safe.\n\n- Flag `tengu_cryptic_horizon`: Not enough to say (read for one account on one subscription tier against v2.1.319; this account: no value returned, anonymous baseline: no value returned, compiled default: off) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.\n- Area: Auto Mode\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 3\/5\n- Scope: individual\n- Heads-up: no"}