In-depth, self-contained guides for the ten applied labs of The Ghost Agent Course by inference.ai. Each lab teaches you to operate Ghost — a ready-to-work AI agent — by prompting it: you don't build an agent from scratch, you direct, debug, guard, and deploy one. Every lab ends with a real artifact you can show.
Each guide is a standalone PDF, sized for students to follow and for proctors to run a session from.
| # | Lab | Difficulty | Guide |
|---|---|---|---|
| 00 | Setup & Debugging Toolkit | Onboarding | |
| 01 | Memory → Living Site | Easiest | |
| 02 | Daily Briefing Agent | Easy+ | |
| 03 | Research → Published Asset | Medium | |
| 04 | Stateful Monitor + Permission Policy | Hard | |
| 05 | Deploy a Shared Operator | Hard |
| # | Lab | Difficulty | Guide |
|---|---|---|---|
| 06 | Data → Decision | Medium | |
| 07 | Growth Engine | Medium | |
| 08 | Inbound Responder | Medium | |
| 09 | Multi-Agent Operation | Hardest (2 sessions) |
Every lab guide contains: learning objectives · why the lab exists · prerequisites · a step-by-step build (with copy-paste prompts) · the failure modes to expect and how to debug them · a deliverable checklist · stretch goals · facilitator notes.
- Prime — load context into Ghost's memory so it knows your world.
- Prompt — give it the task in plain English.
- Break & steer — the first output won't be perfect; debug it with a sharper instruction instead of starting over.
- Crystallize — freeze the working version into a reusable workflow, and wrap it in a permission policy where stakes are real.
The PDFs are built from the HTML fragments in content/ via build.py (uses Playwright/Chromium):
pip install playwright && playwright install chromium
python build.py
The Ghost Agent Course · inference.ai · learn@inference.ai