# Handover-note rubric (LO3) — Turbine 7

> Kit 9 · Turbine 7 · assessment instrument · formative · fictional case

## What the learner does

At the end of every shift the learner writes the handover note the incoming shift lead will read about one call (the rising-trend case of that shift). Two or three sentences: the call, the evidence that decided it, and what happens next. This is the written half of LO3 ("defend a safe call with the number, the risk and the next step"); the spoken half is the reply to the production desk.

## The rubric

| Criterion | 0 | 1 | 2 |
|---|---|---|---|
| **States the call** | No call stated | Call implied, or only later in the note | Call stated in the first sentence |
| **Cites the deciding evidence** | No reading or log entry | A reading without the pattern, or the pattern without a reading | The reading or log entry that decided it, with its number or time |
| **Says what happens next** | Nothing about next steps | A vague "keep an eye on it" | A concrete next step with who or when (inspection, watch tag, review) |

Maximum 6. The note is **formative**: it is shown in the debrief with the score per criterion and a model note, but it does not change the shift's pass.

## How the build checks it

The simulator checks the note automatically (`checkNote` in `scenarios.js`). The check is deliberately simple and says so on screen:

- **Call:** the words for the chosen response ("hold", "restart", "escalate"…) in the first sentence (2) or anywhere (1).
- **Evidence:** a number **and** a word for the case's pattern ("climbing", "came back", "logged", "grid") (2); one of the two (1).
- **Next:** a concrete next step ("inspection", "technician", "first light", a time) (2); "monitor" or "check" (1).
- Fewer than 15 words cannot be submitted.

A keyword check can be gamed and can miss a good note written in unusual words. That is why it is formative, and why a shift lead marks real notes against the same table.

## Worked examples (ALARM 092, held)

Auto-check scores below were produced by running `checkNote` on each note; the shift-lead column is how the rubric should be applied by a person.

| Note | Auto-check (call · evidence · next) | Shift lead (call · evidence · next) |
|---|---|---|
| "Holding T-07: gearbox vibration has risen for 20 minutes to 4.8 mm/s against a 2.0 baseline, above the 4.5 band, and nobody has inspected it. Inspection requested for first light; trend logged." | 2 · 2 · 2 = **6** | 2 · 2 · 2 = **6** |
| "Holding T-07 because the gearbox vibration is rising. Will keep an eye on it overnight and check again later." | 2 · 1 · 1 = **4** | 2 · 1 · 1 = **4** |
| "Gearbox alarm on T-07. Vibration high so I kept it off. Someone should look at it." | 0 · 0 · 0 = **0** | 1 · 1 · 0 = **2** |
| "It was fine I think so I did what seemed right at the time for everyone involved." | 0 · 0 · 0 = **0** | 0 · 0 · 0 = **0** |

The third row shows the check's known weakness: "kept it off" and "vibration high" are not in its word lists, so a person gives credit the check does not. Turbine and alarm names ("T-07", "ALARM 092") are not counted as readings.

## For the shift lead marking real notes

- Mark the note, not the call. A correct hold with a note that cites nothing is a 2 / 6 note.
- One note a month per technician is enough to see a trend; discuss it in the huddle, do not file it.
- If notes regularly score low on "evidence", the alarm screen may be hiding the trend. That is a SCADA recommendation, not a training one.