{"version":"2.1.286","anchor":"auto-mode-classifier-template-guard-project-member-relay-a","canonical_anchor":"auto-mode-classifier-template-guard-project-member-relay-a","heading":"Auto mode's safety checker assembles its rules differently","tier":"internal","area":"Auto Mode","url":"https:\/\/changelogs.core-directive.com\/v\/2.1.286\/e\/auto-mode-classifier-template-guard-project-member-relay-a","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.286","markdown":"### Auto mode's safety checker assembles its rules differently\n\nThe checker behind auto mode now builds its deny rules from one set, records how its template was chosen, and reads tools differently\n\n**Unclear.** It is not clear what switches these changes on or how much of them you would notice.\n\n**What**\n\nAuto mode lets Claude act without asking you each time, with a checker (a classifier) that decides which actions are allowed. Several parts of how that checker is set up changed:\n\n- Its list of rules for blocking actions is now built from one template, and no longer also includes Anthropic's own template.\n\n- It now records which path it took to pick its template, including a stricter fallback and cases where the stricter option was unavailable.\n\n- It records when rows about project members were left out, and why.\n\n- The part that gathers context for it gained options for reading tools and marking which conversation thread something came from.\n\n**Why**\n\nThese change how the auto mode checker's instructions are put together and what it logs.\n\n- Area: Auto Mode\n- Tier: Under the hood\n- Useful: 2\/5\n- Signal: 3\/5"}