Test Instructions

Process

1. Identify the worktree. Absolute path via

   git rev-parse --show-toplevel

   .
2. Skim the diff.

   git diff --stat <base>

   (base resolved via

   references/build-tool-detection.md#base_branch

   ) to know which areas were touched. No deep file reads.
3. Identify the ticket via

   references/issue-tracker-detection.md

   . Skip if none.
4. Pick test commands from

   .turkit.yaml → commands.test

   /

   commands.check

   , or the project's build-tool defaults.
5. Emit the block below using

   references/output-preferences.md

   for operator-facing language/style. No prose around it.

Output template

Test Instructions — <TICKET-ID> <short title>

Worktree:
```bash
cd <absolute-worktree-path>
```

### 1. Automated
```bash
<test/check commands, copy-pasteable, no placeholders>
```

### 2. Live (omit if no runtime to exercise)
```bash
<server start command with env vars if needed>
```
Then:
- <UI step or curl one-liner> → <expected result>

### 3. Edge cases
- [ ] <happy path — one line>
- [ ] <critical edge case>
- [ ] <likely adjacent regression>

### Résumé
<2–3 sentences: what changed, expected behaviour, what to pay attention to during the test>

Rules

• Commands must run as-is. No

  <placeholder>

  except secrets the operator must fill.
• Max ~3 items in Edge cases. Happy path + 1–2 critical edges. If more, you're over-testing.
• Omit "Live" if there's no runtime to exercise (pure refactor, types-only change, docs).
• No "Prerequisites" section, no "Expected result" subsections — fold into the line.
• Apply

  references/output-preferences.md

  for operator-facing language/style. Translate

  Résumé

  naturally when the configured language is not French.