What it does Drop this skill into Claude Code, point it at a repo, and it spits out a self-contained HTML file — no server, no dependencies, works offline. The course walks through how the code actually works using scroll-based modules, animated data flows, side-by-side “code ↔ plain English” panels, and quizzes that test debugging instincts rather than trivia.

The interesting bit The pedagogy is deliberately inverted: build first with AI, understand later with this tool. It assumes you already shipped something (or vibe-coded your way through it) and now need to reverse-engineer your own creation. The quizzes ask practical questions like “A user reports stale data after switching pages — where do you look first?” instead of “What does API stand for?”

Key highlights

• Output is one HTML file; zero runtime dependencies
• Uses exact, unmodified code snippets from the target repo — learners can open the real file and match line for line
• Custom metaphors per concept (auth as a bouncer, rate limiting as nightclub capacity) — no recycled analogies
• Glossary tooltips on hover for technical terms
• “Warm” visual design explicitly defined as not “the typical purple-gradient AI look”

Caveats

• Requires Claude Code and manual installation into ~/.claude/skills/
• The README shows the skill structure but doesn’t clarify how well it handles very large or polyglot codebases
• No mention of customization hooks if the generated course misses the mark

Verdict Worth a look if you’re shipping with AI assistants and tired of cargo-culting code you don’t grasp. Skip it if you already read source fluently or need deep computer science theory — this is practical anatomy, not physiology.