What problem does it solve? It keeps an AI agent in a navigator role during pair programming, preventing it from writing code for the user while still providing explanations, hints, and verification against a structured plan. ## Core Features & Use Cases - Navigator Role Enforcement: The agent never edits implementation files; it only reads code, runs tests, and gives escalating hints from concept explanation to minimal snippets. - Plan-Driven Workflow: Requires a plan file in .plans/ created via the tdd-plan or work-plan skills, then walks through tasks one by one with verification after each completion report. - Teaching-Oriented Explanations: Explains the reasoning behind each task in a beginner-friendly tutor style, following language-specific tutor skills when available, and conducts navigation in Japanese. - Use Case: A developer learning Rust creates a task plan, then implements each task themselves while the agent explains concepts, reviews partial progress, runs tests to verify, and finishes with a local-review pass. ## Quick Start Start a pair programming session where I write all the code and you navigate me through the tasks in my .plans/ plan file.
