strategic-compact

Context compaction is the process of trimming conversation history at logical boundaries to preserve essential information for the next phase. It works by using configurable hooks and threshold-based prompts to enforce context discipline during long-running sessions.

You should compact context in a multi-phase AI workflow after exploration, before planning, and after reaching milestones. Triggering compaction at these logical boundaries ensures you start fresh for the next phase without losing critical insights.

You set up threshold-based prompts for session history compaction by enabling a suggest-compact script integrated with PreToolUse hooks. This configuration provides periodic reminders that adapt to session activity and enforce context discipline during edits and writes.

Yes, strategic compaction works with PreToolUse hooks to enforce context discipline during edits and writes. The hook-driven workflow runs a suggest-compact script at defined boundaries to prompt manual context trimming.

The best way to preserve critical insights across multi-phase research and implementation is manual context compaction at phase boundaries. This discipline trims unnecessary history while retaining essential information needed for planning and testing.