Learn Anything
AI-powered recursive learning system — turns your AI coding assistant into an interactive tutor using the Socratic method and TDD-style exercises.
Generate skill and command files for 30+ AI tools (Claude Code, Cursor, Gemini CLI, Codex, Copilot, Windsurf, etc.), then use slash commands to systematically master any technical topic.
Quick Start
# Initialize in your project (interactive — select your AI tools)
npx learn-anything-cli init
# Or specify tools directly
npx learn-anything-cli init --tools claude
# Chinese terminal output
npx learn-anything-cli init --lang zh-CN
# Or install globally
npm install -g learn-anything-cli
learn-anything initContext7 Integration (Optional)
When running init or update, you'll be prompted to enable Context7 for documentation verification. When enabled, the AI will fetch official documentation and verify its explanations against authoritative sources — improving teaching accuracy.
If you haven't set up Context7 yet, run npx ctx7 setup or visit the Context7 docs for your specific AI tool.
After init, your AI assistant gains five learning commands:
| Command | What it does |
|---|---|
/learn:topic <topic-name> |
Initialize a topic, generate a knowledge map, track progress |
/learn:explain <concept-name> |
Recursive Socratic deep-dive into a concept |
/learn:practice <concept-name> |
TDD-style coding exercises with Socratic feedback |
/learn:review [topic-name] |
Review progress, spaced repetition recommendations |
/learn:status [topic-name] |
Visualize learning state as a knowledge map heatmap |
Learning Workflows
/learn:topic <topic-name> — Initialize a Topic
The AI creates a learning state file (state.json) and generates a knowledge map (knowledge-map.md) via render.mjs, then presents the landscape for you to choose your own path.
/learn:explain <concept-name> — Recursive Deep Dive
The AI assesses your level (beginner → advanced), explains the concept with analogies and code examples, identifies deeper sub-topics, and lets you choose how deep to go. Every session is recorded for spaced repetition.
/learn:practice <concept-name> — TDD-Style Exercises
The AI generates a test-driven exercise at the right difficulty (beginner / intermediate / challenge), provides structured Socratic feedback, and updates your mastery status. Edge cases, security, and code quality are all covered.
/learn:review [topic-name] — Progress Review
Analyzes your learning data: mastery heatmap, spaced repetition priority scoring, concept relationship analysis (blocking / orphan concepts). Generates a personalized next-step plan.
/learn:status [topic-name] — Visualize State
Renders a knowledge map heatmap with status icons, practice counts, and confidence scores for every concept.
Supported AI Tools
Manage, Amazon Q Developer, Antigravity, Auggie, Bob Shell, Claude Code, Cline, Codex, ForgeCode, CodeBuddy Code, Continue, CoStrict, Crush, Cursor, Factory Droid, Gemini CLI, GitHub Copilot, iFlow, Junie, Kilo Code, Kiro, OpenCode, Pi, Qoder, Lingma, Qwen Code, RooCode, Trae, Windsurf, and AGENTS.md-compatible assistants.
# Update skill files to latest version (detects existing tools automatically)
npx learn-anything-cli updateHow It Works
Your Project/
├── .claude/commands/learn/ # Slash commands (learn:topic, learn:explain, ...)
├── .claude/skills/ # Skill files with full workflow instructions
├── .cursor/commands/ # Cursor-specific command format
├── .gemini/commands/learn/ # Gemini TOML-format commands
├── .learn/ # Your learning data (knowledge maps, progress)
│ └── topics/
│ └── javascript/
│ ├── state.json # Learning data (single source of truth)
│ ├── knowledge-map.md # Auto-generated by render.mjs from state.json
│ └── sessions/
└── ...Each AI tool gets tool-appropriate file formats (YAML frontmatter for Claude, TOML for Gemini, etc.) via an adapter pattern.
Development
pnpm install
pnpm build # Compile TypeScript
pnpm test # Run tests
pnpm dev # Watch mode
pnpm dev:cli # Build and run CLI locallyLicense
MIT