context-compression

Design and apply compression strategies for long-running agent sessions. Use when: approaching context limits, reducing token costs, summarizing conversations, optimizing multi-turn interactions, or implementing compaction for agents that run many sequential tasks.