strategic-compact

To prevent context loss in long AI sessions, use strategic manual context compression to prompt compaction at logical milestones, preserving important information before automatic compression randomly discards it.

You should manually compress context during multi-phase AI workflows at logical milestone transitions, such as after completing research or before moving into the implementation and testing phases.

You can trigger manual context compression before a phase transition by enabling the PreToolUse hook, which advises compacting when tool usage reaches a configurable threshold during your workflow.

Yes, you can configure the compaction threshold using the COMPACT_THRESHOLD setting, which determines when the suggest-compact script prompts you to manually compress your session context.

The best way to manage context switches in AI workflows is strategic compaction at milestone transitions, utilizing a dedicated suggest-compact script to guide the compression process and free space.

Automatic context compression causes problems because it triggers randomly, potentially discarding critical information; prompting manual compression at logical milestones ensures important session data is preserved.