What problem does it solve? LLM agents degrade as context grows, but most teams lack the mental models to understand why. This Skill provides the conceptual foundation of context engineering—what context is, how attention mechanics and the U-shaped attention curve constrain agent behavior, and why context quality matters more than quantity—so you can reason about design decisions from first principles before applying specific tactics. ## Core Features & Use Cases - Attention Budget Mental Model: Frames context as a finite attention budget with n-squared pairwise relationships, effective-capacity ceilings below nominal windows, and position-aware placement of critical constraints. - Component Anatomy Guidance: Covers system prompts (altitude calibration, section structure), tool definitions, retrieved documents, message history, and tool outputs with concrete organization principles. - Progressive Disclosure Patterns: Explains three-level progressive disclosure (skill selection, document loading, tool result retention) with a companion Python utility module for token estimation, context building, truncation, and validation. - Use Case: A new contributor joins an agent project and needs to understand why the team's agent loses track of instructions mid-conversation. This Skill explains the U-shaped attention curve and lost-in-the-middle effect, then routes them to the operational skills (context-degradation, context-optimization) that own the fix. ## Quick Start Ask the AI to explain why your agent forgets instructions placed in the middle of a long system prompt and how to restructure it using attention-favored positions.
