{"version":"2.1.283","anchor":"subagents-now-get-a-memory-note-appended-to-their-system-pro","canonical_anchor":"subagents-now-get-a-memory-note-appended-to-their-system-pro","heading":"Subagents now get a memory note, which CLAUDE_CODE_MEMORY_SUBAGENT_APPEND turns off","tier":"use","area":"Memory","url":"https:\/\/changelogs.core-directive.com\/v\/2.1.283\/e\/subagents-now-get-a-memory-note-appended-to-their-system-pro","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.283","markdown":"### Subagents now get a memory note, which `CLAUDE_CODE_MEMORY_SUBAGENT_APPEND` turns off\n\nSubagents that share context now get a memory note added to their instructions; setting `CLAUDE_CODE_MEMORY_SUBAGENT_APPEND` stops it\n\n**Unclear.** What sets the memory mode, and whether some hosts set the new variable themselves, is not established.\n\n**What**\n\nA subagent is a helper agent Claude Code starts to handle part of a task. When a subagent is started without its own separate context, Claude Code now adds a memory note to its system prompt (the standing instructions it receives), unless memory mode is \"none\". Setting the new environment variable `CLAUDE_CODE_MEMORY_SUBAGENT_APPEND` stops this note being added.\n\nSeparately, after a background memory pass writes files, Claude Code now re-reads them and queues a notice that they changed.\n\n**Why**\n\nSubagents now see the same memory context as the main agent. That changes what they know and adds to the size, and so the cost, of their prompts.\n\n- Area: Memory\n- Names: `CLAUDE_CODE_MEMORY_SUBAGENT_APPEND`\n- Tier: Use it now\n- Useful: 4\/5\n- Signal: 3\/5"}