Labs
One folder per lab. Each has INSTRUCTIONS.md (what to do) and checklist.md (what "done" looks like). Every lab
starts with a table that says how long it takes, which preset the shop must be in, and what you need to have done
before. Missed a lab? Each lab says how to catch up, so you can always join the next one.
| Lab | Module | Time | Shop preset |
|---|---|---|---|
| Lab 0 - Arrival | 0 - Arrival & Environment Check | 30 min | clean |
| Lab 2 - Context | 2 - Context Engineering | 25 min | clean |
| Lab 3 - Skills | 3 - Agent Skills | 25 min | clean |
| Lab 4 - RobotCode | 4 - RobotCode for Agents | 25 min | clean |
| Lab 5 - Prompt to green | 5 - Natural Prompt Automation & Debugging | 30 min | clean |
| Lab 6 - MCP | 6 - Robot Framework MCP | 25 min | clean |
| Lab 7 - Hooks and toolbelt | 7 - Hooks, Subagents & the CLI Toolbelt | 20 min | buggy |
| Lab 8 - Healing | 8 - Self-Healing Tests | 12 min | drift_and_bug |
| Lab 9 - CI | 9 - CI & Toolchain Integration | 15 min | clean |
Module 1 (the cold open) and Module 10 (wrap-up) have no lab.
Bonus labs
Self-paced, for after the day. They are not part of the timetable. Bonus 1 and 2 build your own Robot Framework library or tool with an agent, and follow Building libraries and tools with an agent. Bonus 3 has you write subagents of your own for the workshop's suite, and builds on Labs 4 and 7.
| Lab | Chapter | Time | Shop preset |
|---|---|---|---|
| Bonus 1 - A library | Building a library with an agent | about 2 hours | clean |
| Bonus 2 - A tool | Building a tool with an agent | about 60 min | clean |
| Bonus 3 - Subagents | Subagents for debugging and analysis | about 60 min | drift_and_bug |
How a lab works
- Steps are either a command to run or a prompt to give your coding agent. Prompts are quoted in full: paste them as they are, so that everyone's agent gets the same input. Your agent's answer will still differ from your neighbour's. Comparing the two is part of the lesson.
- Presets switch the shop between behaviours:
clean, the drift stages and the planted-bug presets. Apply them only with the shop helper,uv run --no-sync python -m shop preset <name>. It works for the local shop and for your own space on the shared instance alike. The labs that change the preset reset it in their last step. - Claude Code, Codex and GitHub Copilot all work. Where they differ, a step shows a table with one column per agent. The facilitator demonstrates with Claude Code.
- If your agent fails mid-lab, every lab links to a recorded walkthrough of the same lab among the transcripts. Follow along there, and rejoin at the next module.
- After a lab, its reference shows what the rehearsal produced, why it is a good result, and the answers the debrief covers. Open it once you are done: it gives the answers away.
- Stretch goals are for when you finish early. Nothing later depends on them.
Words you have not met before are in the glossary.