strategic-compact

To prevent context loss during multi-step AI workflows, trigger manual compaction at logical boundaries like post-exploration or milestone transitions. This preserves critical context continuity across workflow phases instead of relying on arbitrary automatic compaction.

Strategic context compaction is the practice of manually compressing context at specific workflow boundaries to maintain data continuity. You should trigger it after exploration, before execution, or after completing a major milestone to preserve important plans.

To configure manual compaction suggestions, set up a PreToolUse hook for Edit and Write phases by adding the suggest-compact.sh script to your ~/.claude/settings.json file. You can also define a custom COMPACT_THRESHOLD to control reminder frequency.

Yes, automatic context compaction can cause data loss in AI prompting workflows by arbitrarily compressing context mid-task. Using a hook-based manual approach ensures reminders trigger only at safe logical boundaries, protecting necessary data.

Yes, you can customize the threshold for workflow context management reminders by setting the COMPACT_THRESHOLD variable in your ~/.claude/settings.json file. This allows you to tune the frequency of manual compaction suggestions for your specific process.

You should avoid automatic context compaction in multi-step processes when transitioning between major phases or completing milestones. Arbitrary compression here risks losing critical context, making strategic manual compaction at logical boundaries necessary.