Stem Notes

stem-notes

A Claude Code skill that turns STEM source material (lecture slides, textbook chapters, scanned problem sheets, images, PowerPoint/Word files) into LaTeX study documents: formula/definition sheets, fully worked problem solutions, and summary notes.

How it works

The user is driven through the whole process with multiple-choice questions (AskUserQuestion) — nothing is generated silently. The skill:

1. Inventories the given sources (PDF, PNG/JPG, PPTX/DOCX) and checks whether each PDF is text-based or scanned.
2. Checks for required tooling (poppler, soffice, tectonic/pdflatex) and asks before installing anything.
3. Asks intake questions: which outputs to generate, output location, visual style, and per-output questions (density, scope, depth, ordering, etc.).
4. For a whole book, offers to build a reusable knowledge skill first (see Workflows below) instead of reading every page.
5. Reads every source page/slide/image in full.
6. Extracts figures from PDFs (pdfimages for embedded raster images, page-render + crop for vector diagrams) into a figures/ folder with an index, and asks whether to insert them, catalog them for later, or skip.
7. Solves and verifies every problem in scope, checking non-trivial numbers with a short Python script, and flags any discrepancy with a printed answer instead of silently matching it.
8. Writes ASCII-only .tex files from the templates in templates/.
9. Optionally compiles to PDF (tectonic or latexmk), fixing errors until it compiles cleanly.
10. Reports what was written, what was skipped, and any discrepancies found.

Layout

• SKILL.md — the workflow the skill follows, step by step.
• references/questions.md — the MCQ question bank used for intake.
• references/tooling.md — required tools, install commands/sizes, and the shell commands used to inspect/convert sources.
• references/style-guide.md — layout, labelling, and solution-format rules for each output type.
• templates/formula_sheet.tex — formula/definition sheet preamble.
• templates/solutions.tex — worked-solutions preamble.
• templates/summary_notes.tex — summary-notes preamble.

Workflows

Lecture slides / problem sheets -> notes (the default flow): point the skill at your source files, answer the intake questions, get .tex files (and figures, if any) back.

Book -> knowledge skill -> notes: for a whole textbook, stem-notes asks whether to build a reusable knowledge skill first with book-to-skill, rather than reading every page. This workflow step, and the tool it calls, is the work of virgiliojr94 — credit for it belongs to them, not to stem-notes:

git clone https://github.com/virgiliojr94/book-to-skill ~/.claude/skills/book-to-skill
# or: npx skills add virgiliojr94/book-to-skill

1. Run book-to-skill on the book (e.g. /book-to-skill book.pdf my-book) to get a knowledge skill (SKILL.md, chapters/*.md, cheatsheet.md, glossary.md) — see anderson-flight or callister-materials-science for examples already in this skills folder.
2. Run stem-notes on that generated skill's files to produce formula sheets, worked solutions, or summary notes from it.
3. Keep using the generated book skill afterwards on its own (/my-book <topic>) for questions, worked problems, or more notes — it isn't single-use.

If you already have that book skill installed under a different name, or prefer a different converter skill entirely, stem-notes asks and remembers your choice in the Book skill: line of SKILL.md.

Figures: extracted PDF figures land in latex/figures/ with a figures.md index. You can ask later, e.g. "add figure p12-1 to the summary notes", without re-running extraction.

Output

By default, generated .tex files (and PDFs, if compiled) are written to a latex/ subfolder next to the source files, unless the user chooses otherwise. Extracted figures go in latex/figures/.

License

MIT — see LICENSE.