{"version":"2.1.287","anchor":"memory-selector-prompt-now-requires-a-json-reply","canonical_anchor":"memory-selector-prompt-now-requires-a-json-reply","heading":"Choosing relevant memories now asks for a JSON-only reply","tier":"internal","area":"Memory","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287\/e\/memory-selector-prompt-now-requires-a-json-reply","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287","markdown":"### Choosing relevant memories now asks for a JSON-only reply\n\nThe prompt that picks which saved memories are relevant now asks for a reply made of a single JSON object listing the chosen files\n\n**Unclear.** It is not clear whether a formal output format is also enforced alongside the written instruction.\n\n**What**\n\nClaude Code can keep memories, saved notes stored as files, and it uses a prompt to pick which of them matter for the current task. That prompt now ends with an instruction to reply with only a JSON object (a simple structured text format) of the form `{\"selected_memories\": [\"<filename>\", ...]}`.\n\n**Why**\n\nAsking for one fixed format changes how the choice of memories comes back, which may make that choice more reliable to read.\n\n- Area: Memory\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 1\/5"}