strategic-compact

Manual context compaction prevents context loss during long LLM sessions by prompting compaction at logical task intervals. It tracks tool calls and suggests compacting during natural breaks between exploration, planning, and implementation phases instead of relying on arbitrary auto-compaction.

Manually compact context when transitioning between task phases like moving from research to planning. Phase-aware compaction at these natural breakpoints preserves actionable context, whereas auto-compaction might trigger mid-task and lose critical information.

Phase-aware context management differentiates between exploration, planning, and implementation phases to optimize context retention. It suggests compaction during natural breaks between phases, ensuring relevant context is preserved for subsequent work.

You can configure context compaction thresholds for Claude session management. The skill tracks tool calls and prompts for compaction at configurable thresholds, providing periodic reminders to help manage context proactively throughout your AI workflow.

Context compaction supports Edit and Write tool use hooks for proactive suggestions in Claude. It tracks these tool calls to determine optimal compaction points, ensuring context is preserved during implementation phases involving code editing or file writing.