strategic-compact

To preserve context during long AI assistant sessions, strategic compaction prompts manual context reduction at milestones. This prevents automatic trimming and keeps critical details intact across multi-phase tasks.

The best way to manage context drift is enforcing milestone-based compaction checkpoints. By pruning non-critical context while keeping milestones intact, it enables smoother transitions between research, planning, execution, and review phases.

A pretooluse hook tracks context limits by running a script that monitors tool invocations. Once a configurable threshold is exceeded, it enforces periodic reminders prompting the user to perform manual compaction.

Yes, you can configure the threshold for session persistence reminders. The hook script enforces this configurable threshold to trigger periodic prompts, ensuring you manage context before automatic trimming occurs.

You should perform manual compaction at strategic checkpoints during multi-week workflows or multi-phase tasks. This deliberate approach maintains focus on the current phase without losing prior insights that automatic trimming might discard.