strategic-compact

Compaction prevents degraded model performance by monitoring context window usage and recommending safe context clearing at logical milestones. It ensures tasks, decisions, and uncommitted work are persisted before clearing state, avoiding data loss during long interactive sessions.

Request a context check and compaction recommendation after completing planning, implementation, or triage milestones. This ensures safe handoff guidance by persisting uncommitted work and decisions before recommending a manual clear to prevent data loss.

Yes, safe handoff guidance ensures tasks, decisions, and uncommitted work are persisted before recommending a manual clear. The compaction mechanism never forces clearing state, proactively monitoring context usage and suggesting safe compaction points to prevent data loss.

Yes, context compaction requires visibility into session context usage metrics and milestone signals. The system estimates current context consumption to detect when a session approaches limits, generating safe compaction suggestions to prevent performance degradation.

The best way to maintain model quality is proactive context monitoring combined with milestone-aware compaction. By estimating context consumption and suggesting compaction after implementation milestones, sessions avoid approaching context limits while preserving essential state.