Skip to content

Latest commit

 

History

132 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SkillFlow

An AI-powered curriculum engine for building continuous, human-reviewed technical learning journeys.

SkillFlow turns a structured curriculum into an automated content pipeline — generating, researching, critiquing, reviewing, and publishing technical lessons on a recurring schedule.

The first curriculum powered by SkillFlow is:

Zero to Agentic: Your 105-Day AI Engineer Roadmap


🚀 What is SkillFlow?

Creating a high-quality technical learning series manually is surprisingly difficult.

You need to:

  • Design a coherent curriculum
  • Research each topic
  • Write and explain concepts clearly
  • Ensure technical accuracy
  • Maintain continuity between lessons
  • Create practical examples
  • Review and edit every lesson
  • Publish consistently
  • Track where the learner is in the journey

SkillFlow automates the entire pipeline while keeping a human in the loop.

Instead of treating every article as an isolated piece of content, SkillFlow treats the curriculum as a continuous learning journey.

                    ┌─────────────────────┐
                    │   Fixed Curriculum  │
                    │   105 Day Roadmap   │
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   Select Next Day   │
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   Generate Outline  │
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   Research & Draft  │
                    │   + Web Fact Check  │
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   AI Critique       │
                    │   Accuracy + Clarity│
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   Style & Polish    │
                    └──────────┬──────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │   Human Review      │
                    │   GitHub PR         │
                    └──────────┬──────────┘
                               │
                          Approved?
                         /         \
                       No           Yes
                       │             │
                       ▼             ▼
                    Revise       ┌──────────────┐
                                 │   Publish    │
                                 │ dev.to +     │
                                 │ LinkedIn     │
                                 └──────┬───────┘
                                        │
                                        ▼
                                Next Curriculum Day

🎯 The Core Idea

SkillFlow is built around one principle:

Don't just generate content. Build a learning journey.

Each lesson understands where the learner has been and where they're going.

Every lesson:

  • Recaps the previous day's concept
  • Introduces the current concept
  • Connects it to the broader curriculum
  • Previews the next lesson
  • Uses practical examples
  • Includes code when appropriate
  • Ends with key takeaways
  • Provides a hands-on exercise

This makes the series feel like one continuous course, rather than a collection of unrelated blog posts.


📚 Zero to Agentic

The first curriculum built with SkillFlow is a 105-day journey from NLP fundamentals to production-grade AI agents.

Curriculum

NLP Foundations
      ↓
Text Representation
      ↓
Embeddings
      ↓
Transformers
      ↓
Large Language Models
      ↓
Prompt Engineering
      ↓
Retrieval-Augmented Generation
      ↓
Fine-Tuning
      ↓
AI Agents
      ↓
Agentic Architectures
      ↓
Tool Use & MCP
      ↓
Multi-Agent Systems
      ↓
Evaluation & Observability
      ↓
Production AI Systems

The curriculum is divided into 14 progressive phases, with each phase building on concepts introduced earlier.


🧠 How SkillFlow Works

1. Curriculum Engine

SkillFlow starts with a predefined curriculum.

Rather than asking an LLM:

"What should I teach today?"

the system knows exactly where the learner is in the roadmap.

Day 1 → Day 2 → Day 3 → ... → Day 105

This provides:

  • Deterministic progression
  • Consistent coverage
  • No duplicated topics
  • Controlled learning difficulty
  • A predictable learner journey

2. Multi-Pass Content Generation

A lesson isn't generated in a single LLM call.

SkillFlow uses multiple stages.

Outline

Defines:

  • Learning objective
  • Core concepts
  • Examples
  • Code requirements
  • Previous-day connection
  • Next-day connection

Draft

Generates the initial lesson using the outline and supporting research.

Fact Check

Relevant information is verified using live web research to reduce the risk of outdated or incorrect technical claims.

Critique

The generated lesson is evaluated for:

  • Technical accuracy
  • Pedagogical quality
  • Clarity
  • Logical progression
  • Missing explanations
  • Unnecessary complexity

Style Pass

The final pass makes the lesson:

  • Clear
  • Consistent
  • Teaching-oriented
  • Easy to scan
  • Consistent with the overall series

🔄 Continuity Engine

One of SkillFlow's most important features is lesson continuity.

Traditional AI content generation looks like:

Topic A → Article A

Topic B → Article B

Topic C → Article C

SkillFlow instead creates:

Day 1
  ↓
Day 2 ← Recap Day 1
  ↓
Day 3 ← Recap Day 2
  ↓
Day 4 ← Recap Day 3
  ↓
...
Day 105

Each lesson contains:

Previous Day

What did we learn?

Today's Lesson

What are we learning now?

Next Day

Why does today's concept matter for what's coming next?

This creates a connected curriculum narrative.


👨‍💻 Human-in-the-Loop

SkillFlow is intentionally not fully autonomous.

AI generates the lesson.

A human approves it.

Every lesson goes through a GitHub Pull Request before publication.

AI Generation
      ↓
Automated QA
      ↓
GitHub Pull Request
      ↓
Human Review
      ↓
   Approved?
   /      \
 No        Yes
 ↓          ↓
Revise    Publish

Nothing gets published without human approval.

This provides an important balance:

Automation for scale.
Human judgment for quality.


📤 Automated Publishing

Once a lesson is approved, SkillFlow automatically publishes it to multiple platforms.

dev.to

Lessons are published as part of an official series, providing built-in navigation between lessons.

LinkedIn

Each lesson is transformed into a native LinkedIn post with its own:

  • Headline
  • Teaching structure
  • Key takeaways
  • Appropriate formatting

The same underlying lesson can therefore reach learners across multiple platforms.


⚙️ Fully Automated Scheduling

Once configured, SkillFlow runs unattended.

             WEEKDAY SCHEDULE
                    │
                    ▼
             Find Next Lesson
                    │
                    ▼
            Generate Content
                    │
                    ▼
             Run AI QA Pipeline
                    │
                    ▼
             Create GitHub PR
                    │
                    ▼
              Human Review
                    │
                    ▼
                Approval
                    │
                    ▼
           Publish Automatically
                    │
                    ▼
             Advance Curriculum

The pipeline automatically moves to the next curriculum entry after each successful publication.


🏗️ Architecture

At a high level:

┌──────────────────────────────────────────┐
│              Curriculum                  │
│                                          │
│  105-Day Structured Learning Roadmap     │
└────────────────────┬─────────────────────┘
                     │
                     ▼
┌──────────────────────────────────────────┐
│           Curriculum Engine              │
│                                          │
│  • Day tracking                          │
│  • Phase management                      │
│  • Previous / next context               │
│  • Progression logic                     │
└────────────────────┬─────────────────────┘
                     │
                     ▼
┌──────────────────────────────────────────┐
│          Content Generation              │
│                                          │
│  Outline → Draft → Research → Critique   │
│                         → Style          │
└────────────────────┬─────────────────────┘
                     │
                     ▼
┌──────────────────────────────────────────┐
│             Quality Layer                │
│                                          │
│  • Technical accuracy                    │
│  • Pedagogical quality                   │
│  • Continuity                            │
│  • Formatting                            │
└────────────────────┬─────────────────────┘
                     │
                     ▼
┌──────────────────────────────────────────┐
│            Human Review                  │
│                                          │
│              GitHub PR                   │
└────────────────────┬─────────────────────┘
                     │
                     ▼
┌──────────────────────────────────────────┐
│             Publishing                   │
│                                          │
│          dev.to  +  LinkedIn             │
└──────────────────────────────────────────┘

✨ Key Features

  • 🗺️ Structured curriculum execution
  • 🤖 Multi-stage AI content generation
  • 🔎 Live web-based fact checking
  • 🧠 Context-aware lesson continuity
  • 👨‍💻 Human-in-the-loop approval
  • 🔀 GitHub PR-based workflow
  • 📝 Automatic technical formatting
  • 💻 Code examples and exercises
  • 📚 Sequential learning progression
  • 📢 Multi-platform publishing
  • Scheduled unattended execution
  • 📈 Curriculum progress tracking

🆚 SkillFlow vs Traditional AI Content Generation

Traditional AI Content SkillFlow
Topic-based Curriculum-based
One article at a time Continuous learning journey
Independent articles Context-aware lessons
Single generation pass Multi-stage pipeline
AI-only workflow Human-in-the-loop
Manual publishing Automated publishing
No fixed progression Deterministic progression
Generic content Teaching-oriented curriculum

🔥 Why SkillFlow?

The problem isn't generating another AI article.

The internet already has millions of them.

The harder problem is creating a coherent sequence of lessons that someone can actually follow from beginning to end.

SkillFlow focuses on that problem.

It combines:

Curriculum Design + AI Generation + Research + Quality Control + Human Review + Automation

to turn a static learning roadmap into a continuously delivered learning experience.


🗂️ Project Structure

skillflow/
│
├── curriculum/
│   ├── roadmap
│   └── phases
│
├── pipeline/
│   ├── outline
│   ├── generation
│   ├── research
│   ├── critique
│   └── styling
│
├── continuity/
│   ├── previous_day
│   └── next_day
│
├── publishing/
│   ├── devto
│   └── linkedin
│
├── workflows/
│   └── github_actions
│
├── prompts/
│
├── tests/
│
├── config/
│
└── README.md

Update this section to match the actual repository structure.


🚦 Getting Started

1. Clone the repository

git clone <repository-url>
cd skillflow

2. Install dependencies

pip install -r requirements.txt

3. Configure environment variables

Create a .env file:

LLM_API_KEY=
DEVTO_API_KEY=
LINKEDIN_ACCESS_TOKEN=

Add any additional credentials required by your configured providers.

4. Configure the curriculum

Define or update the curriculum in the curriculum configuration.

Day 1
Day 2
Day 3
...
Day 105

5. Run locally

python main.py

6. Run a specific day

python main.py --day 12

Replace the commands above with the actual CLI commands implemented in the repository.


🧪 Quality Philosophy

SkillFlow follows a simple principle:

AI should accelerate content creation, not replace editorial judgment.

The pipeline therefore uses multiple quality gates before publication.

A lesson should be:

  • Technically correct
  • Understandable to the intended audience
  • Properly connected to the curriculum
  • Useful in practice
  • Consistent with previous lessons
  • Appropriate for the learner's current level

🛣️ Roadmap

Potential future directions include:

  • Personalized learning paths
  • Multiple curriculum templates
  • Learner-specific difficulty adaptation
  • Automated learner progress tracking
  • Interactive exercises
  • Automated quizzes
  • Knowledge assessments
  • Course analytics
  • Additional publishing platforms
  • Multi-language curricula
  • Enterprise learning programs
  • Certification-oriented learning paths

💡 Beyond Zero to Agentic

Zero to Agentic is only the first curriculum.

The long-term vision for SkillFlow is to provide an engine capable of powering many structured technical learning journeys.

                    SkillFlow
                        │
        ┌───────────────┼────────────────┐
        │               │                │
        ▼               ▼                ▼
  Zero to Agentic   Python Mastery   MLOps Journey
        │               │                │
        ▼               ▼                ▼
      105 Days        60 Days          90 Days

The curriculum changes.

The underlying engine doesn't.


📜 License

Add the appropriate license for this repository.


👤 Author

Priyesh Dave

Senior Engineer II – AI/ML

Building at the intersection of AI, automation, and developer education.


⭐ The Vision

Build once. Teach continuously.

SkillFlow is an experiment in using AI not just to generate content, but to engineer an entire learning experience — from the first lesson to the last.

About

AI-powered curriculum engine that turns structured learning roadmaps into continuous, human-reviewed technical content — from generation to automated publishing.

Topics

Resources

Stars

2 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages