strategic-compact

To prevent important context from being lost during long AI sessions, apply manual memory compaction at major milestones to discard irrelevant logs and preserve the decision record. This ensures crucial context survives memory refresh.

Compact memory to preserve design decisions at strategic moments such as design completion, major milestones, or after debugging. Compacting at these points discards nonessential logs while keeping the decision record intact for future use.

Export a reference snapshot like SUMMARY.md to keep decisions accessible after memory refresh. This documents the compacted context, ensuring decisions and essential context remain available even after the thread is cleared.

The best way to compact context without losing planning details is to ensure planning is fixed before compacting. Once planning is stable, compact the context to maintain focus on implementation while preserving the decision record.

Memory optimization works for end-to-end workflows after debugging by proposing, performing, and documenting the compacted context. This ensures decisions survive the memory refresh and the workflow continues with a clean context.