compaction-skill

Conversation summarization compacts verbose history into a structured, token-efficient recap. It produces a five-section summary that replaces existing memory while appending new messages to maintain workflow continuity across extended multi-turn assistant phases.

Summarizing long conversations for state preservation is best handled by generating a structured five-section recap: Task Overview, Current State, Important Discoveries, Next Steps, and Context to Preserve. This captures decisions, failures, and critical details to avoid repeating effort.

Compacting AI memory during extended multi-turn workflows involves triggering a summarization process as token thresholds approach. This compresses prior messages into a concise summary designed for seamless handoff into subsequent conversation turns.

Memory compaction should be run when the token budget is approaching its limit to avoid losing workflow continuity. It converts verbose conversation history into a concise summary that retains the information needed to continue effectively.

Yes, conversation summarization captures specific critical details and constraints. The Context to Preserve section records what was decided, why, what worked or failed, and specific values, IDs, preferences, and constraints to avoid repeating effort.