# Yuki's Session — Schema

Flat CSV in `yuki-session.csv`. Two logical sections in one file, disambiguated by the `section` column.

## Columns

| Column | Type | Description |
| --- | --- | --- |
| `section` | text | Either `reroll` (an iteration row) or `diagnostic` (a scorecard row). |
| `index` | integer | Iteration number (1..4) for reroll rows; diagnostic id (1..5) for diagnostic rows. |
| `label` | text (quoted) | Iteration prompt for reroll rows; diagnostic name for diagnostic rows. |
| `detail` | text (quoted) | Copilot outcome for reroll rows; evidence for the diagnostic firing for diagnostic rows. |
| `fires_or_verified` | boolean | `verified` value for reroll rows (always `false` in this scenario); `fires` value for diagnostic rows (always `true` in this scenario). |

The `label` and `detail` fields are CSV-safe JSON strings so quotes and commas inside the prompt or evidence do not break parsers.

## Row inventory

- 4 `reroll` rows (indexed 1–4).
- 5 `diagnostic` rows (indexed 1–5).

Total: 9 rows.

## Deliberate design choices

Unlike the numeric-scenario datasets (Priya, Rohan), Yuki&rsquo;s dataset has no arithmetic to reconcile. Its purpose is to lock the <em>shape</em> of the vibe-charting failure: four prompts that never harden into criteria, zero verified iterations, five diagnostics that all fire. Every row is a locked fact that keeps the print figure and the chapter narrative in lockstep.

The one designed-in tension is that Chapter 13&rsquo;s <em>Running the diagnostics on Yuki&rsquo;s session</em> section walks the diagnostics as if the reader is watching them fire live. The CSV carries the same evidence strings the chapter prose uses so the two stay synchronised.

## License

Released for teaching use alongside *The Defensible Decision*. Use freely.