learning-workspace-template

A workspace for systematically learning open-source projects, powered by AI agent skills that work with both Cursor and Claude Code.

How it works

This workspace uses agent skills to guide you through learning any GitHub project in about 2 hours. The process is structured into 5 stages: onboard, prerequisites, build/test/run, hands-on exercises, and ecosystem mapping.

Quick start

Cursor

• Open learning.code-workspace in Cursor
• Ask: **"I want to learn "**
• The learn-project skill kicks in and walks you through 5 stages
• Notes and exercises appear in notes/ /
• The cloned repo lives in repos/ /

Claude Code

• cd into the workspace directory
• Run claude to start a session — it picks up CLAUDE.md automatically
• Ask: **"I want to learn "**
• Claude reads the skill file and follows the same 5-stage workflow

What the skill produces

For each project you learn, you get:

AI tool compatibility

Both tools get the same skills and conventions — just through different discovery mechanisms.

Prerequisites

• Cursor IDE and/or Claude Code
• Git
• GitHub CLI (gh) — for fetching repo metadata
• Language-specific tools as needed (Node.js, Python, etc.)

Tips

• Use agent-driven mode ("learn this project for me") to have the AI do the heavy

lifting — it clones, builds, tests, and writes the guide while you review.