What problem does it solve? AI coding agents hallucinate APIs, ignore project conventions, and degrade in quality during long sessions when they receive too little, too much, or stale context. This Skill provides a structured method for deciding what information the agent sees, when it sees it, and how it is ordered. ## Core Features & Use Cases - Five-Level Context Hierarchy: Organizes context from persistent rules files (CLAUDE.md, AGENTS.md, .cursorrules) down to per-iteration error output and conversation history. - Context Budget Management: Trims failed attempts and verbose tool output starting at 75% window capacity, compresses exploration into one-line summaries, and positions task-critical content last to counter the lost-in-the-middle effect. - Confusion Management: Surfaces spec-versus-code conflicts and missing requirements as explicit options instead of silently guessing, plus an inline planning pattern for multi-step tasks. - Use Case: When an agent starts inventing imports that do not exist mid-project, apply this Skill to write a rules file covering tech stack, commands, conventions, and boundaries, then reload only the relevant spec section and source files for the current task. ## Quick Start Ask the agent to audit the current session's context setup and create a CLAUDE.md rules file covering the project's tech stack, commands, conventions, and boundaries.
