AURA Core

Agentic Unified Reinforcement for Academics

A multi-agent AI framework for designing, developing, and quality-assuring educational content at scale.

The Problem

Educational course development is fragmented across multiple specialties—curriculum design, content writing, presentation creation, assessment design, and quality assurance. This leads to:

Misalignment between learning outcomes, activities, and assessments
Inconsistent quality and accessibility standards
Difficulty coordinating specialized expertise
No systematic quality gates before delivery

The Solution

AURA provides an orchestrated multi-agent system where 7 specialized AI agents work in sequence, each with defined responsibilities, built-in quality gates, and evidence-based pedagogical frameworks.

┌─────────────────────────────────────────────────────────────────┐
│                    AURA Execution Flow                          │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│   Orchestrator ──► Content Developer ──► Presentation Creator   │
│        │                                        │               │
│        │                                        ▼               │
│        │              ◄── QA Educator ◄── Assessment Architect  │
│        │                      │                                 │
│        │                      ▼                                 │
│        └──────────────► Quality Gate ──► Ready for Delivery     │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

Installation

# Run directly in your project
npx aura-edu-core

# Or install globally
npm install -g aura-edu-core

This copies the .aura-core/ framework and installs Claude Code slash commands.

The 7 Specialized Agents

Agent	Role	Focus
Educational Orchestrator	Coordinates end-to-end workflows	Progress tracking, stage transitions
Educational Designer	Structures courses and outcomes	Bloom's alignment, scaffolding
Content Developer	Drafts lessons, examples, exercises	Practice-first, worked examples
Presentation Creator	Creates slides and visuals	Dual-coding, accessibility
Assessment Architect	Designs tests and rubrics	Validity, reliability, Bloom levels
Learning Curve Specialist	Optimizes difficulty progression	Pacing, spacing, scaffolding
QA Educator	Quality assurance gate	O→A→T traceability, accessibility

Core Workflows

1. Course Planning

Research → Outcomes Map → Curve Analysis → Assessment Strategy → QA Gate

2. Course Execution (per lesson)

Shard Course → Draft Lesson → Create Slides → Build Assessment → QA Review

Artifact Standard Paths

All outputs follow consistent file paths:

docs/courses/{course_slug}.md              # Course specification
docs/outcomes/{course_slug}-map.md         # O→A→T matrix
docs/lessons/{course_slug}/{lesson_id}.md  # Lesson scripts
docs/slides/{course_slug}/{lesson_id}.md   # Presentations
docs/assessments/{course_slug}/{id}.md     # Assessments
docs/rubrics/{course_slug}/{id}.md         # Rubrics
docs/qa/gates/{course_slug}-{id}.yml       # QA decisions

Project Structure

.aura-core/
├── agents/          # 7 agent role definitions
├── workflows/       # Planning & execution orchestration
├── templates/       # YAML/MD blueprints for artifacts
├── checklists/      # Quality & accessibility criteria
├── tasks/           # Executable task guides
├── data/            # Pedagogical frameworks & knowledge base
└── docs/            # User guides & documentation

Embedded Pedagogical Frameworks

Bloom's Taxonomy — 6 cognitive levels with measurable verbs
Assessment Levels — L1 (micro-formative) to L4 (capstone)
Instructional Patterns — Worked examples, retrieval practice, dual coding
Cognitive Load Theory — Chunking, signaling, modality principles
Accessibility Standards — WCAG AA equivalent baseline
O→A→T Traceability — Outcomes → Activities → Tests alignment

Quality Gates

Every lesson passes through a QA gate before delivery:

Status	Meaning
PASS	Ready for delivery
CONCERNS	Minor issues noted, can proceed
FAIL	Must address issues before delivery
WAIVED	Exception granted with justification

Claude Code Slash Commands

After installation, use these commands in Claude Code:

Command	Agent	Description
`/orchestrator`	Educational Orchestrator	Start workflows, shard courses
`/designer`	Educational Designer	Design courses, create outcome maps
`/content`	Content Developer	Draft lessons, examples, exercises
`/slides`	Presentation Creator	Create slide decks and visuals
`/assessment`	Assessment Architect	Design tests and rubrics
`/curve`	Learning Curve Specialist	Analyze difficulty progression
`/qa`	QA Educator	Quality review, gate decisions

Example Usage

/orchestrator start the python-basics course workflow
/content develop lesson L01 on variables
/slides create presentation for L01
/assessment design quiz for L01
/qa review L01 and issue gate decision

Key Principles

Outcomes First — Everything traces back to learning outcomes
Agent Specialization — Each agent has a focused responsibility
Progressive Scaffolding — Content builds systematically
Feedback Loops — Continuous quality checks
Accessibility by Default — Inclusive design from the start

Requirements

Node.js 18+
npm or npx

License

MIT

Built for educators who want AI assistance without sacrificing pedagogical quality.

AlenPaunov/aura-core

README