What problem does it solve? Hierarchical multi-agent systems waste tokens by replaying the full orchestrator trajectory into every worker call, while summarization adds latency and loses information. This Skill explains how to transfer orchestrator state to workers at the representation level by compacting the worker model's KV cache instead of re-serializing text. ## Core Features & Use Cases - Task-guided KV compaction: Score trajectory positions using queries from the current worker prompt, keeping only what the worker will actually attend to. - Shared global mask with MAD thresholding: Aggregate attention scores across layers and heads into one mask, using a robust median + tau * MAD threshold instead of per-head top-k. - Decision framework: Compare Latent Briefing against prefix caching, summarization, and RAG to pick the right mechanism for your bottleneck. - Use Case: In a recursive orchestrator-worker system doing long-document QA, compact the growing orchestrator trajectory before each worker call to cut worker tokens while preserving task-relevant reasoning state. ## Quick Start Ask the AI to design a KV cache compaction strategy for your orchestrator-worker agent system using task-guided Attention Matching.
