Context Management

This gives you proactive token budget control instead of scrambling after compaction hits. Check your current usage, create snapshots at 70-85% capacity, and restore context at three detail levels (essential is ~250 tokens, comprehensive is ~1,250). Version 3.0 adds burn rate tracking so you get predictions like "you'll hit 70% in ~15 tool uses" and health indicators for your statusline. The priority system keeps requirements and decisions while archiving verbose logs. Best move is checking status regularly during long sessions and snapshotting before major operations. Works alongside the PreCompact hook but catches problems earlier when you still have room to maneuver.