/claude-docs-course

name: claude-docs-course
description: >
  Generate an interactive, self-contained HTML course on any Claude documentation
  topic. Use this skill when the user wants to create an interactive course,
  tutorial, or deep-dive walkthrough about a Claude feature, API concept, SDK
  pattern, or prompt engineering technique. Triggers on: "create a course about",
  "interactive tutorial for", "teach me about X interactively", "make a course on",
  "I'd like a course", "/docs --course", "/docs course", or when the user accepts
  the post-docs course prompt. Produces a stunning single-page HTML file with
  scroll-based navigation, animated visualizations, quizzes, and code translations
  drawn from official Claude documentation.

Docs-to-Course

Transform any Claude documentation topic into a stunning, interactive single-page HTML course. The output is a single self-contained HTML file (no dependencies except Google Fonts) that teaches the topic through scroll-based modules, animated visualizations, embedded quizzes, and plain-English translations of real API examples and configuration snippets from the official docs.

First-Run Welcome

When the skill is first triggered and the user hasn't specified a topic yet, introduce yourself:

> **I can turn any Claude topic into an interactive course — visual explanations, animated diagrams, and hands-on quizzes, all in a single HTML file.** > > Tell me a topic: > - **A Claude Code feature** — e.g., "create a course on hooks" > - **An API concept** — e.g., "make a course about tool use" > - **An SDK pattern** — e.g., "interactive tutorial for Agent SDK sessions" > - **A technique** — e.g., "teach me about extended thinking interactively" > > I'll read through the official documentation, design a learning arc, and generate a beautiful single-page HTML course with animated protocol diagrams, code explanations, and interactive quizzes. Opens right in your browser — no setup needed.

If the user just came from a `/docs` response, the topic is already known — skip the welcome and start building.

Who This Is For

The target learner is a **developer building with Claude** — someone who has general technical literacy but wants to deeply understand a specific Claude feature, API concept, or SDK pattern before using it in their project.

**Assume general technical literacy.** The learner knows what APIs, functions, JSON, and HTTP requests are. They don't need basic programming concepts explained. But they DO need Claude-specific concepts, patterns, and terminology explained thoroughly — tokens, context windows, tool_use blocks, stop sequences, system prompts, streaming events, hook matchers, MCP servers, etc.

**Their goals are practical:**

• **Master a Claude feature** before using it in production — understand not just the API surface, but the mental model, edge cases, and best practices
• **Make confident architectural decisions** — know which Claude feature to use for a given problem, understand tradeoffs
• **Debug effectively** — when something doesn't work as expected, know where to look and what to check
• **Stay current** — Claude's capabilities evolve fast; deep understanding beats surface-level familiarity
• **Communicate precisely** — use the correct terminology when discussing Claude integrations with teammates

**They are NOT beginners at programming.** They're experienced developers who are new (or deepening) in a specific Claude feature. The course should respect their time and intelligence while making the Claude-specific material stick.

Why This Approach Works

Documentation is comprehensive but flat — it lists features, parameters, and examples without a learning arc. This skill transforms documentation into a **structured learning experience** with visual explanations, interactive elements, and progressive disclosure.

The learner already has context: they've used Claude, they've called the API, they may have built with some features already. The course meets them where they are: "You've been passing messages to the API — here's what's actually happening under the hood, and here's how to unlock the advanced patterns."

Every module answers **"when and why would I use this?"** before "how does it work?" The answer is always practical: *because this pattern solves a real problem you'll encounter when building with Claude.*

The single-file constraint is intentional: one HTML file means zero setup, instant sharing with teammates, works offline, and forces tight design decisions.

---

The Process (4 Phases)

Phase 1: Topic Discovery

Before writing course HTML, deeply understand the topic by reading all relevant documentation. Thoroughness here pays off — the more you understand, the better the course.

**How to find documentation files:**

Use the search scripts (they read the manifest + index, so they see every page whether or not it is cached):

1. **Content search** — for questions or compound topics:

   bash ~/.claude-code-docs/plugin/skills/claude-docs/scripts/content-search.sh "<keyword1>" "<keyword2>"

2. **Fuzzy search** — for approximate names:

   bash ~/.claude-code-docs/plugin/skills/claude-docs/scripts/fuzzy-search.sh "<query>"

3. **By category** — list a whole product's pages:

   jq -r '.pages[] | select(.category=="claude_code") | .filename' ~/.claude-code-docs/paths_manifest.json

**Reading a page:** pages are cached at `~/.claude-code-docs/cache/<filename>`. If a file isn't there yet, fetch it first, then read it:

~/.claude-code-docs/plugin/scripts/fetch-docs.sh get "<filename>"

(Fetching several? Run `~/.claude-code-docs/plugin/scripts/fetch-docs.sh sync` once instead.)

**What to read:**

• Read the top 5-8 matching docs for the topic (fetching any that aren't cached)
• Read 1-2 adjacent/related docs for context (e.g., if the topic is "hooks", also skim the skills and settings docs)
• Cap at 10 docs total to avoid context exhaustion

**Wh