# New Course

> Adds a new course to the spec-build-lab. Use this skill ANY time the user wants to register, scaffold, set up, add, create, init, or start tracking a NAMED COURSE in this repo — including phrasings like "scaffold a new course called X", "set up a course folder for X", "make a new course called X", "add the X course", "init the X course", "let's start tracking the X course", "I want to add a new course", "/new-course X", or any user message naming a course they want to begin. This skill copies templates/course/ to courses/<name>/ and nothing else; it does NOT read materials, write spec.md, or run a build. If the user asks anything about an EXISTING course, do NOT use this skill. Always run from the repo root.

- Skill: `https-deeplearning-ai/new-course` (Agent Skill)
- Install (CLI): `npx skillmds@latest add https-deeplearning-ai/new-course`
- Raw SKILL.md: https://api.skillmd.com/api/skills/https-deeplearning-ai/new-course/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: https-deeplearning-ai (https://skillmd.com/u/https-deeplearning-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/https-deeplearning-ai/new-course

---


# new-course

Scaffold `courses/<name>/` from `templates/course/`. This skill ONLY copies
the template; it does not read materials and does not generate `spec.md`.

## Steps

1. **Verify cwd is the repo root.** Confirm both paths exist relative to cwd:
   - `templates/course/`
   - `.claude/skills/new-course/`
   If either is missing, refuse with: "Run /new-course from the repo root —
   the folder containing templates/ and .claude/." Stop.
2. **Require the course name.** If `$ARGUMENTS` is empty (or the user did not
   provide a name in their message), ask: "What should the course be called?
   It becomes the folder name under courses/." Stop until they answer.
3. **Refuse on collision.** If `courses/$ARGUMENTS/` exists, refuse with:
   "courses/$ARGUMENTS/ already exists. Pick a different name or remove the
   existing folder first." Stop.
4. **Copy the template:**
   ```bash
   cp -R templates/course/ "courses/$ARGUMENTS/"
   ```
5. **Report and hand off.** Tell the user, in this order:
   1. "Scaffolded courses/$ARGUMENTS/."
   2. "Get notebooks + transcripts from
      https://course-context-lab.vercel.app — that's where this lab's
      course materials are produced. Download them yourself; I won't
      WebFetch from there."
   3. "Tip: once they're downloaded, drag the files from Finder into
      this terminal window. The terminal pastes their absolute paths,
      and I can `cp` them into
      courses/$ARGUMENTS/materials/notebooks/ and
      courses/$ARGUMENTS/materials/transcripts/ — saves tokens versus
      having me traverse your filesystem."
   4. "Then `cd courses/$ARGUMENTS` and run /generate-spec when the
      materials are in place — say 'go ahead' when you're ready."
   5. "Optional extra: if you also want the course's real code, find the
      course at https://github.com/deeplearningai-eng/courses and run
      `/ingest-course-repo <course-folder-URL>` before /generate-spec — it
      adds the `.ipynb` files and `helper.py` under
      courses/$ARGUMENTS/materials/notebooks/from-repo/. This is an addition
      to step 2, not a replacement: the notebooks `.md` and the transcripts
      still come from the site."

## Don't

- Don't read or write `materials/`, `spec.md`, `builds/`, or `evals/`.
- Don't `cd` for the user — they choose when to move into the course folder.
- Don't run from inside an existing course folder. The session is anchored
  at the repo root.
- Don't WebFetch from `course-context-lab.vercel.app` — point the user at
  it; let them download manually. The agent doesn't pull course content
  from the web.