{"version":"2.1.323","anchor":"compaction-summaries-told-how-to-treat-elided-tool-results-a","canonical_anchor":"compaction-summaries-told-how-to-treat-elided-tool-results-a","heading":"Conversation summaries get rules for left-out tool output and older text","tier":"notice","area":"Compaction","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.323\/e\/compaction-summaries-told-how-to-treat-elided-tool-results-a","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.323","markdown":"### Conversation summaries get rules for left-out tool output and older text\n\nAfter a long conversation is condensed, Claude is told to re-read earlier text with a tool rather than rewrite it from memory\n\n**What**\n\nWhen a conversation gets long, Claude Code can condense it into a summary to free up room. This is called compaction. Two new instructions now guide Claude after that happens:\n\n- Tool results left out entirely: Claude should only report that the tool call was made, not guess what it returned. A tool call is one action Claude takes, such as reading a file or running a command.\n\n- Text from before compaction: that text is no longer in Claude's working memory, so Claude should read it again with a tool, using the conversation id `current`, instead of reproducing it from memory.\n\n**Why**\n\nThese instructions change what Claude says about earlier parts of a conversation once it has been condensed. They steer it away from restating content it can no longer actually see.\n\n- Area: Compaction\n- Tier: You'll notice\n- Useful: 1\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}