{"version":"2.1.321","anchor":"system-prompt-new-guidance-sections-wired-in","canonical_anchor":"system-prompt-new-guidance-sections-wired-in","heading":"System prompt gains finishing-work, memory and privacy guidance sections","tier":"internal","area":"System Prompt","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.321\/e\/system-prompt-new-guidance-sections-wired-in","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.321","markdown":"### System prompt gains finishing-work, memory and privacy guidance sections\n\nThe system prompt can now include new guidance sections on finishing work, memory and notes, and user privacy, and its research limit is no longer always empty\n\n**Unclear.** Which models or settings turn these sections on is not known.\n\n**What**\n\nThe system prompt is the set of standing instructions Claude Code gives the model before your conversation. The code that builds it gains several new sections, plus one changed setting:\n\n- `userPrivacy` is a new entry.\n\n- `hasFinishingWorkGuidance` controls a new section about finishing work, listed as `finishing_work`.\n\n- `hasMemoryGuidance` controls a new section about memory, listed as `notes_and_feedback`.\n\n- `researchLimit` is no longer always empty (`null`). Before, it was hard-wired to nothing.\n\nThe `finishing_work` and `notes_and_feedback` names also appear in the list Claude Code uses to track which prompt sections differ between runs.\n\n**Why**\n\nThese entries let the system prompt include extra instructions on how Claude wraps up work, how it handles memory and notes, and user privacy. When they are included, they can change how Claude behaves near the end of a task and what it keeps in memory. The findings do not show the wording of these sections or when they are included.\n\n- Area: System Prompt\n- Tier: Under the hood\n- Useful: 2\/5\n- Signal: 3\/5\n- Scope: individual\n- Heads-up: no"}