What problem does it solve? Understanding an unfamiliar codebase requires hours of manual reading. This Skill automates codebase exploration by spawning parallel agents that analyze architecture, code snippets, and usage patterns, then writes structured documentation you can reference later. ## Core Features & Use Cases - Parallel Agent Exploration: Launch 1, 3, or 5 Haiku agents depending on depth mode (--fast, default, --deep) to analyze different aspects of a repository simultaneously. - Organized Documentation Output: Generates time-prefixed markdown files (ARCHITECTURE, CODE-SNIPPETS, QUICK-REFERENCE, TESTING, API-SURFACE) in date folders, plus a hub index file linking all sessions. - Repository Management: Clones repos via ghq, maintains origin symlinks, and tracks learned repos in an .origins manifest with an --init restore mode. - Use Case: You discover an interesting open-source library on GitHub. Run the learn command with the repo URL to get architecture diagrams, code examples, and a quick reference guide written to your knowledge base within minutes. ## Quick Start Ask the AI to learn a repository by providing its GitHub URL, for example by saying learn https://github.com/owner/repo with the default three-agent mode.
