make-lecture-kit — turn any topic into an easy, visual, interactive study kit

Give it a topic, a list of topics, your notes or study material, or PDFs / slide decks — and your AI assistant produces two things into an output/ folder:

• companion.pdf — a professionally typeset study companion in plain, easy English: an everyday analogy before every idea, real-life examples, and every worked example walked through step by step (LaTeX + figures).
• lecture.html — a complete, very interactive lecture: the whole topic as a story-driven page where you play with every idea and watch the intuition appear.

No slides needed. With just a topic name, the assistant designs the syllabus, the analogies and the worked examples itself. With material, it covers all of it and fills in the steps, intuition and examples the material skips. The output is generic and unbranded — use it for any subject, any course, any learner.

No API keys. You install nothing. Your own agent session writes the source, makes the figures, and compiles the PDF for you. It works first-class on all three major assistants — Claude, OpenAI Codex, and Google Jules (plus Cursor, Gemini CLI, and the like) because it ships both open skill formats: a plain SKILL.md (Claude and Codex auto-discover it) and an AGENTS.md at the repo root (Jules, Codex, Cursor, Gemini CLI, and the wider agents.md ecosystem read it natively). Nothing is platform-specific — standard LaTeX plus standard-library Python — and the companion PDF compiles wherever TeX is available (most agent sandboxes have it built in).

---

Install (pick your assistant)

Everything lives in one repo: https://github.com/saurabh1604/make-lecture-kit. Cloning is best — it makes staying current a single command (see the next section). Prefer not to use git? Download the ZIP from that page and unzip it wherever the steps below say to clone.

Claude Code

git clone https://github.com/saurabh1604/make-lecture-kit ~/.claude/skills/make-lecture-kit

Claude Code auto-discovers any skill in ~/.claude/skills/. (For one project only, clone into /.claude/skills/ instead.) Then ask: “Use make-lecture-kit on the topic eigenvectors.”

OpenAI Codex

git clone https://github.com/saurabh1604/make-lecture-kit ~/.agents/skills/make-lecture-kit

Codex auto-discovers SKILL.md skills in ~/.agents/skills/ (and in .agents/skills/ inside a repo). List them with /skills, invoke explicitly with $make-lecture-kit, or just say: “Use the make-lecture-kit skill on the topic backpropagation.” (No-git option: in Codex run $skill-installer and point it at the repo.)

Google Jules

Jules runs on a GitHub repo and reads AGENTS.md at the repo root automatically. Two ways to use it:

• Fastest: in Jules, connect the saurabh1604/make-lecture-kit repo, then ask “Use make-lecture-kit to make a companion PDF and an interactive lecture for .” Jules reads AGENTS.md → SKILL.md and follows it.
• In your own repo: drop the make-lecture-kit/ folder into your project (or copy its AGENTS.md to your repo root) and point Jules at that repo.

Claude Cowork

Add the make-lecture-kit folder as a skill via Settings → Capabilities (the “add skill” flow), or save it from the shared .skill file. Then just ask for a lecture kit in chat.

Cursor / Gemini CLI / any other agent

The kit is platform-neutral. Clone it or drop the folder into your project; the root AGENTS.md (and SKILL.md) tell the agent exactly what to do. Then ask: “use make-lecture-kit on .”

Or don’t install at all

Drop the make-lecture-kit folder into a chat with any agent and say: “Use this skill to make a companion PDF and an interactive lecture for .”

---

Staying updated

The kit improves over time (every change bumps VERSION and adds a CHANGELOG.md line). Pulling the latest takes one command — and your own work in output/ is never touched.

• Cloned it (Claude Code, Codex, Cursor): pull in the folder you cloned into —

  cd ~/.claude/skills/make-lecture-kit      # or ~/.agents/skills/make-lecture-kit
  git pull                                   # or:  python3 scripts/update.py
  

  python3 scripts/update.py --check previews what’s new without changing anything.
• Downloaded the ZIP: run python3 scripts/update.py — it reads the latest VERSION straight from GitHub and installs the new kit over your copy (keeping output/).
• Google Jules: nothing to do — Jules re-clones the repo for every task, so it always runs the newest version. (If you forked the repo, click Sync fork on GitHub now and then.)
• Claude Cowork: re-add the newer folder/.skill, or use the Claude Code clone above to get the one-command updater.

---

Use

Tell your assistant any of these (more in references/prompts.md):

• “Use make-lecture-kit on the topic eigenvectors.”
• “Use make-lecture-kit on these topics: mean, median, mode, variance.”
• “Use make-lecture-kit on my attached notes / PDF / slides.”
• “Explain photosynthesis simply for a Class 10 student, with worked examples.”

You’ll get a new folder output// containing companion.pdf (the typeset study companion) and lecture.html (the complete interactive lecture), alongside the companion.tex source and the figures/ used to build it.

Behind the scenes, your assistant reads SKILL.md and follows its workflow (work out what kind of input it got, build a concept inventory, write every idea with an analogy first, work every example in full step by step, build the page and the PDF, then run the quality gates). You don’t do any of that — just give it a topic or your material.

About the PDF: it compiles instantly in Claude Cowork and Codex sandboxes (TeX Live is built in). On a bare machine without TeX, the assistant will say so plainly, leave the ready-to-compile companion.tex + figures, and tell you the one-line TinyTeX install to finish the job.

---

What’s inside

make-lecture-kit/
├─ START_HERE.md                read this first (1-minute setup)
├─ README.md                    this file
├─ SKILL.md                     the instructions your AI follows (Claude, Codex)
├─ AGENTS.md                    same kit, read natively by Jules / Codex / Cursor
├─ VERSION                      current version (the updater compares this)
├─ CHANGELOG.md                 what changed in each version
├─ update_source.txt            where scripts/update.py pulls updates from
├─ publish.sh                   maintainers only: self-check, commit and push a release
├─ templates/
│  ├─ companion.tex             LaTeX study-companion template (→ PDF)
│  ├─ lecture.html              complete interactive lecture template
├─ references/
│  ├─ plain_language.md         the easy-English rulebook (both linters enforce it)
│  ├─ quality_rubric.md         the quality bar + ship checklist (both)
│  ├─ source_modes.md           any input (topic / list / notes / PDF / slides) → concept inventory
│  ├─ companion_style.md        how to write the companion
│  ├─ lecture_style.md          how to build the complete interactive lecture
│  ├─ intuition_playbook.md     analogies, mental models, real-world links
│  ├─ prompts.md                copy-paste prompts for learners
│  └─ upgrading.md              the improve-the-kit loop (for maintainers)
├─ scripts/
│  ├─ figstyle.py               matplotlib house style + reusable plotters
│  ├─ build_pdf.py              run figures + compile companion.tex → PDF
│  ├─ _plain_language.py        shared word lists used by both linters
│  ├─ lint.py                   lecture HTML quality gate (readability, no-overflow, keyless)
│  ├─ lint_tex.py               companion PDF-source language + layout gate
│  ├─ selfcheck.py              verify the whole kit is healthy
│  └─ update.py                 pull the latest kit (keeps your output/)
├─ examples/
│  ├─ sample_companion.tex      a finished example to show the quality bar
│  └─ figures/
│     └─ example_normal_curve.py  copy-adaptable house-style figure script
└─ output/                      your generated kits land here (the skill
                                creates one subfolder per topic — don't
                                edit this folder by hand)

The companion uses LaTeX + matplotlib (already present in Cowork/Codex sandboxes); the helper scripts are standard-library Python. Nothing phones home, nothing needs an API key.

---

The promise

Built around seven rules: easy words, simple things kept simple (the most intuitive way in, right-sized — never complicated), relatable analogies in one story-world, step-by-step math intuition, generous fully-solved examples, interactions that uncover the intuition, and lead with the whole idea in one line (headline titles, open by doing, close on a pocket card) — with a hard no-clutter, no-overflow guarantee so text never runs off the page or screen.