# Course Specification: Shifting from Workgentic to Agentic Workflow
**Slug**: `workgentic-to-agentic`  
**Task Key**: `workgentic-to-agentic`  
**Creator**: Buzz & BuildBotsBetter  
**Target Audience**: All AI Agents in Fleet (claude_cli, grok_bot, grok_build, chatgpt_codex, cursor_hobby, qwen_code, antigravity_cli)

---

## 1. Course Definition
- **Title**: Shifting from Workgentic to Agentic Workflow
- **Subtitle**: Stop writing every line. Start directing, testing, debugging, and supervising the AI that writes it.
- **Core Thesis**: Code Writer → Code Director / Technical Overseer.

---

## 2. Opening Visual & Conceptual Anchor
```text
PROGRAMMING EVOLUTION:
Books → Stack Overflow → Search → ChatGPT → Copy/Paste Code → Copy/Paste Errors → AI-Assisted Coding → CLI / Cursor / Coding Agents → AGENTIC WORKFLOW
```

### Contrast Table
- Workgentic: I write code, I debug code, I know what changed, I run tests, I fix errors, I know how it works, I am the implementer.
- Agentic: I direct the agent, I provide debugging tools, I inspect what changed, I require the agent to test, I have the agent diagnose/fix, I verify that it works, I am the overseer.
- Rule: *Agentic doesn't mean "trust the AI." It means change where your effort goes.*

---

## 3. Curriculum Modules (1 to 12)
- **Module 1 — You're Not Coding the Same Way Anymore**: Old Programmer vs AI-Assisted vs Agentic Programmer.
- **Module 2 — What Actually Changes?**: The new control loop (`Human → Objective → Agent → Code/Tools/Commands → Environment → Evidence → Agent → Human Oversight`).
- **Module 3 — Prompting Becomes Delegation**: Don't tell what code to write. Specify Objective, Scope, Constraints, Verification, and Evidence.
- **Module 4 — The Trust Problem**: `Trust ≠ Verification`. Valid evidence criteria.
- **Module 5 — Give the Agent Eyes**: The Agent Debugging Toolkit & Observability.
- **Module 6 — Don't Just Give the Agent Tools**: Build the surrounding environment (`PROJECT/` structure).
- **Module 7 — The Agentic Development Loop**: `DEFINE → INSPECT → PLAN → EXECUTE → TEST → OBSERVE → DIAGNOSE → CORRECT → VERIFY → REVIEW ↺`.
- **Module 8 — When the Agent Crashes**: 8-step recovery procedure instead of manual takeover.
- **Module 9 — Stop Being the Human Stack Trace**: System-level self-investigation by the agent.
- **Module 10 — Human-in-the-Loop**: 3 levels of delegation (Assisted, Supervised, Autonomous).
- **Module 11 — Your New Job Description**: The Agentic Developer responsibilities.
- **Module 12 — Build Your First Agentic Workflow (Capstone)**: Building an agentic project environment with verification scripts and diagnostic harness.

---

## 4. Capstone & Final Test: "You Are the CTO Now"
- **Scenario**: The agent has just broken the application.
- **Test**: Fix it without writing the fix yourself. Oversee diagnosis, require evidence, verify repair.

---

## 5. Course Core Principle
> **Don't ask: "Can AI write the code?"**  
> **Ask: "Can I build a workflow that lets AI produce, test, debug, and prove the code is correct?"**