compact-discipline

To prevent context drift in long AI sessions, you need smart context management to prune history. Running /compact at strategic thresholds summarizes and refreshes the session, preventing degraded responses.

You should run /compact to reclaim tokens when context usage nears thresholds after reading large files, during long explorations, or before beginning a new logical unit to prevent degraded responses.

To preserve critical decisions when pruning context history, maintain a checkpoint before discarding data. This checkpointing ensures you can resume work seamlessly from the updated state after compaction.

The best way to manage context usage after exploring large datasets is triggering /compact based on session progress. This prunes history and reclaims tokens without losing critical decisions made during exploration.

Yes, you can resume your workflow after running /compact by following encoded operational requirements. The system provides resume guidance to continue from the updated checkpoint, ensuring continuity across agent teams.