AI-Courses

A collection of HTML-based courses in "For Dummies" style, built with Claude Code. Each course targets a specific audience and covers a focused topic — from AI fundamentals to on-device Android ML to personal productivity agents.

Project Structure

AI-Courses/
├── CLAUDE.md                          # Project-wide instructions
├── index.html                         # Full course catalog
├── .claude/commands/                  # Root-level skills (scaffold + audit/refresh/append/mark-proofread)
│
├── CourseName/                        # Each course follows this layout:
│   ├── CLAUDE.md                      # Course description & chapter details
│   ├── .claude/commands/
│   │   ├── generate-next-chapter.md   # Chapter generation command
│   │   └── create-memory-map.md       # Visual summary generation command
│   └── chapters/
│       ├── index.html                 # Landing page with chapter navigation
│       ├── assets/style.css           # Shared CSS (dark/light theme)
│       ├── 01-topic.html              # Numbered chapter files
│       ├── 02-topic.html
│       └── maps/                      # Visual memory maps
│           ├── 01-topic-map.html
│           └── 02-topic-map.html

Workflows

All workflows use Claude Code slash commands. Make sure you're in the correct directory before running them.

1. Create a New Course

From the root directory (AI-Courses/):

/scaffold-new-course

Provide a description of what the course should cover, who it's for, and any specific topics. Claude will scaffold the full course structure.

2. Generate Chapters

Navigate inside the course directory first, then run:

/generate-next-chapter

This generates the next chapter in sequence based on the course's CLAUDE.md plan.

3. Create a Memory Map

Memory maps are single-page visual summaries of a chapter — concept cards, flow diagrams, quizzes, and key takeaways. From inside the course directory:

/create-memory-map 01

Replace 01 with the chapter number you want to summarize.

4. Maintain an Existing Course

Once a course is published, the world keeps moving — model versions ship, vendor docs reshuffle, new techniques appear. Four root-level skills (run from the root AI-Courses/ directory) keep courses fresh:

/audit-course <course-id>           # produce a refresh+append plan in <course>/UPDATES.md
/refresh-next-chapter <course-id>   # apply one refresh from the queue
/append-next-chapter <course-id>    # generate one new chapter from the queue
/mark-proofread <course-id> [color] # toggle a "Quality Checked" ribbon

The flow is audit → refresh/append → mark-proofread:

• /audit-course is read-only. It re-verifies every load-bearing claim with fresh WebSearch and writes a UPDATES.md plan into the course — two queues (chapters to refresh, new chapters to append) plus the source URLs it consulted. It never edits chapters itself.
• /refresh-next-chapter pops one item from the refresh queue, re-grounds it with fresh WebSearch, applies the edits (honoring Preserve ranges from UPDATES.md), regenerates the chapter's memory map, and ticks the checkbox. Re-invoke to advance.
• /append-next-chapter pops one item from the append queue, re-verifies the topic is still warranted, generates the new chapter, wires it into CLAUDE.md / chapters/index.html / the previous chapter's footer, bumps the chapter count in the root catalog, and (if the course uses them) generates a memory map. Re-invoke to advance.
• /mark-proofread toggles a green or orange "QUALITY CHECKED" ribbon on the course card in index.html and on the chapter index hero — used after you've personally read a course end-to-end.

All three maintenance commands re-run WebSearch at execution time (not just at audit time), so a stale UPDATES.md won't quietly land bad facts.

5. Recommended Flow: Full Course Generation

For generating an entire course (chapters + memory maps) efficiently, use this batch workflow:

1. cd into the course directory
2. Enter plan mode (once for the whole session)
3. Run /generate-next-chapter
4. Run /create-memory-map XX (where XX is the chapter just generated)
5. Repeat steps 3–4 three to four times
6. Run /compact to free up context
7. Repeat steps 3–6 until all chapters are done

Why plan mode? It helps Claude maintain a coherent plan across chapters and keeps generation consistent.

Why compact? Each chapter is 400–600 lines of HTML. After a few chapters, compacting the conversation prevents hitting context limits and keeps generation fast.