Генерация учебных задач с ключом ответов

learning

scaffold-exercises is a Claude Code skill that generates graded problem sets for teaching, emitting two separate files: a clean student copy containing only problems and a solution key with worked answers plus a short "why this matters" rationale for every problem. It covers three problem types — analytical (derivation, optimization, comparative statics), empirical (estimation and interpretation on a dataset), and coding (runnable R or Stata implementations) — and, when no dataset is provided, simulates one with a fixed date seed so the answer key stays deterministic. Before generation the skill reads source materials such as lecture `.tex` or `.qmd` files and produces a Pre-Flight Report that confirms the topic, learning objectives, and per-type problem counts. Flags `--difficulty` (intro / core / advanced), `--count`, and `--types` give fine-grained control over scope and depth. It is intended for instructors preparing homework, practice sets, or graded assignments with a ready answer key — not for grading submissions or building timed exam infrastructure.