What problem does it solve? AI agents often degrade unpredictably as conversations grow, tool outputs pile up, and instructions get lost in long prompts. This Skill provides the foundational mental models and concrete thresholds needed to design, debug, and optimize how context is assembled for language model agents. ## Core Features & Use Cases - Context anatomy guidance: Covers system prompts, tool definitions, retrieved documents, message history, and tool outputs with actionable engineering patterns for each component. - Attention and budgeting mechanics: Explains the lost-in-the-middle effect, effective capacity limits (60-70% of nominal windows), and compaction triggers at 70-80% utilization. - Progressive disclosure patterns: Describes three-level loading strategies for skills, documents, and tool results to keep context high-signal. - Use Case: When an agent's reasoning quality collapses after 30 tool calls, use this Skill to diagnose message-history bloat and apply observation masking and compaction strategies. ## Quick Start Ask the AI to explain why your agent loses track of instructions in long conversations and how to fix its context budget.
