Optimizes token usage and session health by suggesting context compaction at logical task boundaries.
The Strategic Context Compactor skill enhances Claude Code sessions by replacing unpredictable auto-compaction with intelligent, logic-driven prompts. By monitoring tool usage thresholds—specifically for file editing and writing—this skill alerts you when it is time to prune the context window. This ensures that essential implementation details are preserved while clearing out stale data, research logs, and debugging cycles that could otherwise lead to model hallucinations or inefficient token consumption during long-running tasks.
Key Features
01Periodic reminders to maintain session health during long tasks
0261 GitHub stars
03Configurable thresholds for initial compaction suggestions
04Intelligent tool usage tracking for Edit and Write calls
05Token optimization for improved response accuracy
06Manual control over context pruning at logical milestones
Use Cases
01Pruning research context after a technical plan is finalized but before implementation begins
02Refreshing the context window when switching between different feature modules in a single session
03Clearing extensive error logs and debugging history after a successful bug fix
