strategic-compact

Manage Claude AI token limits during long tasks by triggering manual context compaction at logical workflow intervals, such as between research and implementation phases, to prevent losing valuable information from arbitrary auto-compaction.

Context compaction is the process of summarizing or reducing conversation history to preserve context window space. Use it at natural workflow transitions, like after completing research, to keep implementation plans clear and accessible.

Prevent arbitrary auto-compaction from disrupting your workflow by proactively suggesting manual `/compact` commands based on tracked tool usage thresholds, ensuring context is preserved at logical breakpoints rather than random limits.

Strategic context compaction works specifically for long sessions approaching token limits and multi-phase tasks. It tracks tool calls and suggests compaction at configurable thresholds, such as after 50 tool calls, with periodic reminders.

Auto-compaction loses valuable information during coding tasks because it triggers arbitrarily at token limits rather than at logical phase transitions. Manual compaction at workflow intervals ensures relevant research context and implementation plans remain accessible.

The best way to preserve context through task phases in Claude AI is phase-aware context management. Prompt compaction at natural breaks, such as before starting implementation after research, to maintain a clear and accessible project plan.