What problem does it solve?
Auto-compaction triggers at arbitrary points during long AI sessions, often mid-task, leading to loss of critical context like partial implementation state, variable names, and recent reasoning. This Skill eliminates that risk by prompting you to run manual /compact only at logical workflow boundaries where context loss is minimal.
Core Features & Use Cases
- Tool Call Tracking: Monitors the number of tool invocations in your session to identify natural compaction checkpoints.
- Phase-Aware Guidance: Includes a decision matrix to help you determine when compaction is appropriate based on your current task phase (e.g., after planning, after debugging, before switching tasks).
- Best Practice Rules: Provides clear guidelines for what context persists through compaction and what is lost, so you can save important information to files or memory before compacting.
- Use Case: For a long development session spanning research, implementation, and testing, this Skill ensures you compact only after finalizing your implementation plan, not mid-coding where you would lose track of partial changes.
Quick Start
Enable the strategic-compact skill to receive prompts to run /compact only at logical task boundaries instead of relying on arbitrary auto-compaction.