Run sheet

Facilitator material. The day, module by module: what the shop must be in, what you show, what participants do, and what to do when it goes wrong or runs late. Print it. Times are CET, as in the master preparation document.

The day

Time	Module	Shop preset	You demonstrate	Participants do	Fallback	When it runs late
09:00-09:30	0 - Arrival & Environment Check (30 min)	clean	Nothing yet: welcome, the poll "what should an agent take off your plate first?"	Lab 0	Co-host breakout with the triage playbook; shared instance; pairing	Stragglers continue in the breakout during Module 1
09:30-09:55	1 - Cold Open, the Agentic Shift & the Maturity Ladder (25 min)	clean on your machine; the cold open switches it	The cold open, then the ladder and the standing rule	Listen; glossary at hand	The recorded cold open	Shorten the talk, never the cold open
09:55-10:45	2 - Context Engineering (50 min: 15 / 25 / 10)	clean	An AGENTS.md written live; the same prompt before and after	Lab 2	Steps 2-4 need only an editor; transcript	Debrief down to 5 minutes
10:45-11:00	Break
11:00-11:50	3 - Agent Skills (50 min: 15 / 25 / 10)	clean	A SKILL.md taken apart; trigger and no-trigger	Lab 3	Transcript	Skip the stretch; debrief 5 minutes
11:50-12:35	4 - RobotCode for Agents (45 min: 15 / 25 / 5)	clean	discover, libdoc, robot-debug on a test you break live, the REPL misfire	Lab 4	Steps 3-4 run by hand; transcript	Skip the REPL step (6)
12:35-13:20	Lunch
13:20-14:05	5 - Natural Prompt Automation & Debugging (45 min: 10 / 30 / 5)	clean	/opsx:propose on a slice outside the lab (WEB-002_AC-8, the price range)	Lab 5, pair review in breakouts	Transcript; the pair review works on its proposal	Announce the two-criteria slice; skip the archive step (7). The rehearsal's agent needed 14 minutes to propose and 7 to apply four criteria
14:05-14:50	6 - Robot Framework MCP (45 min: 15 / 25 / 5)	clean	Your Module 5 demo test rebuilt stepwise	Lab 6	Transcript	Skip the cost step (6)
14:50-15:05	Break
15:05-15:45	7 - Hooks, Subagents & the CLI Toolbelt (40 min: 15 / 20 / 5)	buggy, applied by each participant in step 4	Hooks firing; gh issue create from a real run; 10 minutes on prompt injection	Lab 7	Steps 3-5, 7-8 need no agent; transcript	Cut stretch B (Jira)
15:45-16:15	8 - Self-Healing Tests (30 min: 15 / 12 / 3)	drift_and_bug, applied by each participant in step 1	A live healing run, and its report triaged (answers)	Lab 8	The recorded report	First cut: demo only, skip the lab
16:15-16:50	9 - CI & Toolchain Integration (35 min: 15 / 15 / 5)	clean (CI starts its own shop)	A pushed break explained in a pull request on your fork	Lab 9	Transcript	Cut the stretch (heal suggestions)
16:50-17:00	10 - Wrap-Up, Roadmap & Q&A (10 min)	-	The ladder once more; the Monday plan: AGENTS.md, then the RobotCode plugin and RF Agent Skills, then one CI triage step, then the bonus chapters for building your own library or tool	Questions	-	-

Buffer lives in debriefs and stretch goals, never in lab time.

Cuts

When the day runs late, cut in this order, and only as far as needed:

1. the Module 8 lab: show Module 8 as a demo;
2. the Module 9 stretch goal;
3. Module 7's stretch B, the Jira skill.

A beginner-heavy room (you know at registration): run Module 8 as a demo only from the start, and give its 12 lab minutes to Module 2.

Presets for the room

Every participant switches their own shop, local or in their own space, with the shop helper. Say the command out loud and put it in the chat at the start of Modules 7 and 8, and the reset at their end:

uv run --no-sync python -m shop preset buggy          # Module 7, step 4
uv run --no-sync python -m shop preset drift_and_bug  # Module 8, step 1
uv run --no-sync python -m shop reset                 # the end of Modules 7 and 8
uv run --no-sync python -m shop status                # what holds right now

Presets compose: buggy keeps a drifted layout if one is active. The reset always returns to clean. A participant whose suite fails more than the two broken tests outside Modules 7 and 8 has a preset left on: status, then reset.

Healing keys (Module 8)

Decide before the setup email whether this workshop hands out healing keys. Without keys, Module 8 works with the recorded report and the agentic stretch goal, and nobody needs anything extra.

If you hand them out:

1. Create one key per participant at an OpenAI-compatible provider, each with a hard spending cap and an expiry at the end of the day. One healing run under drift_and_bug takes about 20,000 tokens; a cap for ten runs is plenty.
2. Send each key in a private chat message on the day. Never in a shared document, a slide or the repository.
3. Participants put HEAL_MODEL, HEAL_BASE_URL and HEAL_API_KEY into .env (SETUP.md, Healing API key). setup-check then shows Healing endpoint as passed.
4. Revoke all keys after the workshop.

Never one shared key for the whole room: it cannot be capped per person, and it leaks the moment someone pastes it.

Module 5: the whiteboard

During the debrief, build the group's review checklist on the shared whiteboard. Participants leave with it. Start from this skeleton, and let the room fill in what their pair reviews found:

Question to ask of an agent's plan	Found in our reviews
Does it cover exactly the slice, no more?
Are its specs in suite/*, naming each criterion?
Does every test name start with its criterion?
Do locators live in resources, built on the stable contract?
Are quoted texts from the spec used as they are?
Does every test create its own state?
What could the agent not know without seeing the page?

The last row is the bridge to Module 6.