strategic-compact

To preserve context in long-running AI assistant sessions, you can trigger manual context compaction at logical workflow boundaries like phase transitions. This approach retains essential plans and decisions while discarding irrelevant history, maintaining thread continuity.

Strategic context compaction manages context bleed by prompting manual compaction at logical milestones rather than arbitrary limits. It preserves essential context and maintains coherence across multi-phase projects by persisting only relevant information at phase boundaries.

You set up manual compaction prompts by installing and configuring a PreToolUse hook with a configurable threshold. This triggers periodic reminders at key transitions, guiding you on when to compact and what context to persist.

You should compact context during multi-phase projects at key transitions, such as moving from research to planning or planning to implementation. This deliberate compaction at phase boundaries keeps only relevant context and balances thoroughness with momentum.

Auto-compaction relies on arbitrary limits, while manual compaction at logical boundaries is better for rapid task switching. Strategic compaction allows human-guided persistence of essential context, preserving thread continuity that arbitrary auto-compaction might lose.

Yes, you can configure the compaction reminder cadence using a configurable threshold. This allows you to balance thoroughness with momentum by adjusting when the strategic compaction prompts appear during your multi-phase tasks.