skill tutorial · scenario
A retro that survives its own minutes
Most retros die at the parking lot. The Start-Stop-Continue folder is built against that — one behavior per card, never a person, no card without an owner and a date — and the skill wires the same discipline to measurement: six items before, six after, a 30-day follow-up that counts behaviors, not feelings.
Install and brief the agent
the whole room kit in one loop
curl -LO https://workshopskill.com/en/activities/start-stop-continue/files/start-stop-continue.zip unzip start-stop-continue.zip > retro, one team of 6, 45 minutes, last sprint was rough the skill scales: card stock per person, three verbs on the wall, 30-minute core, the folder's debrief; metadata duration is the truth, your calendar is the negotiation
The session, by the folder's rules
Silent card-writing first (the file is explicit: talk after writing, never during), one behavior per card, the grouping pass, then the three-column dot vote. The failure modes are in the debrief with rescues — the dominant storyteller, the diplomatic blank wall, the action items that read like mission statements. The agent read them into your prep; now you own them in the room.
Measurement, three beats
same items, three moments
pre (before writing cards): 6 Likert items — honesty, behavior-over-person framing, commitment quality post (after dot vote): the same 6, the deltas land in the FormLM report automatically day30 (follow-up): the folder's 3-item check — did the owned, dated cards actually happen?
The report you hand upward
Deltas on the framing items say whether the retro culture moved; day-30 counts say whether commitments did. That pairing is the whole Level-2-to-Level-3 bridge in miniature — the training-evaluation folder does it at classroom scale, this one at team scale, and neither rounds commitment percentages up past what the cards can support.
keep going
questions
Do I need FormLM for the 30-day check?
No — the follow-up runs on the paper cards themselves: owner, date, done-or-not. The live lane just automates the tally and keeps the history.
Can the agent run the retro for me?
It prepares, scales, prints and measures. The room is the human's — the folder's facilitation notes exist precisely because a model cannot hold the silence after someone's name comes up.
Team of more than 8?
The folder draws the line at one team of 4–8 per session — split into two sessions and keep the cards shared; the skill's group_size says so before anyone improvises.