Course Builder

Course Builder — Online Course Full Production Pipeline

Collaboratively produce an online course's curriculum, lesson plans, quizzes, and hands-on labs through an agent team, all in one pass.

Execution Mode

Agent Team — Five agents communicate directly via SendMessage and cross-validate each other's work.

Agent Roster

Workflow

Phase 1: Preparation (Performed Directly by the Orchestrator)

1. Extract from user input:
   • Course Topic: Subject/field the course covers
   • Target Learner: Beginner/Intermediate/Advanced, background knowledge
   • Course Scale: Total learning time, number of modules
   • Lab Environment (optional): Tools, languages, platforms to use
   • Existing Files (optional): Curriculum, lesson plans, etc.
2. Create the _workspace/ directory at the project root
3. Organize input and save to _workspace/00_input.md
4. If existing files are available, copy to _workspace/ and skip the corresponding phase
5. Determine execution mode based on scope

Phase 2: Team Assembly and Execution

Tasks 2a (lessons), 2b (quizzes), and 2c (labs) run in parallel. All depend only on Task 1 (curriculum).

Inter-agent communication flow:

• curriculum-designer completes -> delivers per-lesson objectives and concepts to content-writer; Bloom's level ratios to quiz-maker; lab environment and scenarios to lab-designer
• content-writer completes -> delivers key concepts and examples to quiz-maker (as item material); lesson content to lab-designer (for lab alignment)
• course-reviewer cross-validates all deliverables. On RED Must Fix findings, sends revision requests -> rework -> re-validate (up to 2 cycles)

Phase 3: Integration and Final Deliverables

1. Verify all files in _workspace/
2. Confirm all RED Must Fix items have been addressed
3. Report final summary to user

Execution Modes by Scope

Using Existing Files: If the user provides a curriculum, lesson plans, etc., copy to _workspace/ and skip the corresponding agent.

Data Transfer Protocol

File naming convention: {order_number}_{deliverable}.{extension}

Error Handling

Test Scenarios

Normal Flow

Prompt: "Design a 10-hour online course for Python web development beginners, focused on the Flask framework." Expected Result:

• Curriculum: 3-4 modules, 3-5 lessons each, Bloom's-based learning objectives
• Lesson plans: Per-lesson teaching flow, slide outlines, code examples
• Quizzes: Formative (3-5 items per lesson) + Summative (10 items per module)
• Labs: Per-lesson coding labs + Capstone project (simple web app)
• Review: Learning objective coverage matrix with full mapping

Existing File Flow

Prompt: "Create quizzes and lab assignments from this curriculum" + attached curriculum file Expected Result:

• Copy existing curriculum to _workspace/01_curriculum.md
• Combined quiz + lab mode: deploy quiz-maker + lab-designer + reviewer
• Skip curriculum-designer and content-writer

Error Flow

Prompt: "Just design a quick curriculum on data analysis" Expected Result:

• Switch to curriculum mode (curriculum-designer + reviewer)
• Target learner unclear; default to "beginner (non-specialist)" and note in report
• Review report notes "Lesson plans/quizzes/labs not generated"

Agent Extension Skills