course-generate

Generate a complete course — learning objectives, syllabus, slides (LaTeX), scripts, and assessments — using the ADDIE multi-agent pipeline. Use when the user asks to "create a course on X", "generate teaching materials for X", "build an undergraduate/graduate course on X", or similar end-to-end course-authoring requests.

Course Generate (Full ADDIE Pipeline)

Runs the complete multi-agent course generation pipeline:

  • Phase 1 — Foundation deliberations (learning objectives, resource assessment, target audience, syllabus, assessment planning, final project)
  • Phase 2 — Per-chapter development (slides, scripts, assessments via SlidesDeliberation)
  • Phase 3 — Evaluation (Program Chair + Test Student review)

This is a long-running task — typically 20–60 minutes depending on model and number of chapters. Confirm scope with the user before launching.

When to invoke this skill

  • User asks to generate a course / teaching materials / curriculum
  • User provides a topic/subject and wants end-to-end output
  • User wants to reproduce the paper's pipeline on new content

Do not invoke for:

  • Just converting existing LaTeX → PPTX → use latex-to-pptx
  • Only evaluating existing slides → use slide-evaluate
  • Optimizing an existing slide chapter → (future: slide-optimize skill)

Prerequisites

  • pip install instructional-agents
  • OPENAI_API_KEY set in the environment
  • Disk space ~10–50 MB per course

Usage

python3 "${CLAUDE_PLUGIN_ROOT}/skills/course-generate/scripts/generate.py" \
  --course "<course name>" \
  [--model <model_name>] \
  [--exp-name <tag>] \
  [--catalog <catalog_name>] \
  [--copilot <copilot_name>]

Arguments

ArgRequiredDescription
--courseyesCourse name/topic, e.g. "Reinforcement Learning"
--modelnoLLM model (default: gpt-4o-mini)
--exp-namenoSubdirectory under exp/ (default: test)
--catalognoPre-loaded reference catalog name
--copilotnoCopilot mode with interactive feedback

Example

python3 "${CLAUDE_PLUGIN_ROOT}/skills/course-generate/scripts/generate.py" \
  --course "Introduction to Reinforcement Learning" \
  --model gpt-4o \
  --exp-name rl_undergrad_2026

Output structure

exp/<exp-name>/
├── learning_objectives.md
├── resource_assessment.md
├── target_audience.md
├── syllabus.md
├── assessment_planning.md
├── final_project.md
├── chapter_1/
│   ├── slides.tex
│   ├── slides.pdf          (if compiled)
│   ├── script.md
│   └── assessment.md
├── chapter_2/...
└── evaluation/
    ├── program_chair_review.md
    └── test_student_review.md

Follow-up skills

After generation you typically want to run:

  • latex-compile → produce PDFs
  • latex-to-pptx → produce editable PowerPoint (per chapter)
  • slide-evaluate → re-evaluate with different rubrics

Citation

This is the full pipeline from Instructional Agents (arXiv:2508.19611). Please cite if used in published work.