What problem does it solve?
This Skill prevents slow responses and runaway token costs by helping you control when session context should be compacted or cleared between unrelated tasks or long loops.
Core Features & Use Cases
- Token-cost-aware context resetting: Uses clear thresholds (e.g., ~100k tokens, >4 hours, switching tasks) and explains why caching doesn’t eliminate per-turn output cost.
- Mid-task cleanup with /compact: Shows when to compact during the same task while preserving the plan and relevant state.
- Task-boundary handover with /clear: Guides writing a handover document to disk before clearing so the next session can resume correctly.
- Loop and subagent cost hygiene: Provides checkpointing and routing guidance to reduce context bloat during ralph/autopilot/background loops and manage subagent-driven costs.
Quick Start
If your session has exceeded about 100k tokens or you’re switching to a new task, write a short handover to disk and then start the new phase with a clean context.