TestSAFE

AI automation skills specifically designed for Unity

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: unity-test
description: Run Unity Test Runner operations asynchronously
---

> **Before calling any skill in this module:** if you are about to call a skill with parameters guessed from its name or description, STOP — read this file (or fetch its schema via `GET /skills/recommend?includeSchema=true`) first. If you already have the parameter definitions from recommend/schema, you may proceed straight to dryRun.

## Triggers
- Running EditMode/PlayMode tests
- Discovering or listing tests
- Polling async test results
- Scaffolding test files
- 运行 EditMode/PlayMode 测试、发现或列出测试、轮询异步测试结果、生成测试文件

# Test Skills

Run and manage Unity tests.

## Operating Mode

- **Approval**: 只读 skill(`test_get_result` / `test_list` / `test_discover_get_result` / `test_get_last_result` / `test_list_categories` / `test_smoke_skills` / `test_get_summary`,标 `SkillMode.SemiAuto`)直接执行;执行/发现/创建型 skill(`test_run` / `test_run_by_name` / `test_discover_start` / `test_cancel` / `test_create_editmode` / `test_create_playmode`,默认 `SkillMode.FullAuto`)需用户 grant,grant 后服务端一步执行返结果(job 立即排进队列)。
- **Auto / Bypass**: 直接执行。
- **本模块有 4 个 NeverInSemi skill**(按 `IsForbiddenInSemi` 自动判定):
  - `MayEnterPlayMode = true`: `test_run`、`test_run_by_name`
  - `MayTriggerReload = true`: `test_create_editmode`、`test_create_playmode`(同时标 `MutatesAssets = true`)

  Approval 模式下这 4 个返 `MODE_FORBIDDEN`,仅 Bypass 或 Allowlist 命中可绕过。注意:`test_run(testMode="PlayMode")` / `test_run_by_name` 会让 Unity TestRunner 切入 PlayMode;`test_create_editmode` / `test_create_playmode` 落盘新的 .cs 文件后会触发 Domain Reload。
- **异步约定**:`test_run` / `test_run_by_name` / `test_discover_start` / `test_create_*` 立即返回 `jobId`;用 `test_get_result(jobId)` / `test_discover_get_result(jobId)` 轮询;Unity TestRunner 串行化,**正在跑测试时不要再起第二个 `test_run`**。

**DO NOT** (common hallucinations):
- `test_run_all` does not exist → use `test_run` or `test_run_by_name`
- `test_create_template` does not exist → use `test_create_editmode` or `test_create_playmode`
- `test_get_status` does not exist → use `test_get_result` with `jobId` from test run
- Test skills are async — they return a `jobId`, poll with `test_get_result(jobId)`
- Unity Test Runner is serialized here: do not start a second `test_run` while another test job is still active
- Prefer `unity_skills.get_skills(category="Test")` or `GET /skills/schema` for exact signatures instead of guessing from memory

**Routing**:
- For compile error checking → use `debug` module's `debug_check_compilation`
- For test script creation → `test_create_editmode` / `test_create_playmode`, then modify via `script` module
- For broad regression probes across many skills → `test_smoke_skills`, which uses transient probes to avoid polluting workflow/batch persistence

## Skills

### `test_list`
List available tests via Unity Test Runner async discovery. **Returns `pendingDiscovery=true` + `discoveryJobId` on first call (cache miss)** — poll `test_discover_get_result(jobId)` then retry `test_list`.
**Parameters:**
- `testMode` (string, optional): EditMode or PlayMode. Default: EditMode.
- `limit` (int, optional): Max tests to list. Default: 100.

**Returns:** `{ success, testMode, count, total, truncated, tests, pendingDiscovery, discoveryJobId, discoveryStatus }` — here `count` is how many came back and `total` is how many were discovered, `truncated: true` means raise `limit`. (`test_discover_get_result` uses `count` for the total instead — see the warning there.)

### `test_run`
Run Unity tests asynchronously. Returns a `jobId` immediately; poll with `test_get_result(jobId)`.
**Parameters:**
- `testMode` (string, optional): EditMode or PlayMode. Default: EditMode.
**Returns:** `{ success, status, jobId, kind, testMode, filter, message }`

### `test_get_result`
Get the result of a test run.
**Parameters:**
- `jobId` (string, required): Job ID from `test_run` / `test_run_by_name`.

**Returns:** `{ success, jobId, status, totalTests, passedTests, failedTests, skippedTests, inconclusiveTests, otherTests, failedTestNames, failedTestDetails, elapsedSeconds, resultSummary, error }` — each failure detail includes `name`, `resultState`, `message`, `stackTrace`, `durationSeconds`, and `output`.

### `test_cancel`
Cancel a running test job if supported (Unity TestRunner has no hard cancel — best-effort).
**Parameters:**
- `jobId` (string, required): Job ID to cancel.

**Returns:** `{ success, jobId, status, cancelled, note, warnings }`

### `test_discover_start`
Start asynchronous Unity Test Runner discovery and return a discovery `jobId`. Use this directly when you want explicit control over discovery; otherwise `test_list` / `test_list_categories` will trigger it on cache miss.

| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| testMode | string | No | EditMode | EditMode or PlayMode |

**Returns:** `{ success, status, jobId, kind, testMode, message }`

### `test_discover_get_result`
Get the result of an asynchronous Unity Test Runner discovery job.

| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| jobId | string | Yes | - | Discovery job ID |
| limit | int | No | 100 | Max tests in the response; **the default silently caps a larger suite** |

**Returns:** `{ success, jobId, status, testMode, discoveryMode, count, returned, truncated, tests, error }`

> **Here `count` is the total and `returned` is what you got.** `count` is how many tests the discovery found, `returned` is how many are in `tests` after `limit` was applied, `truncated: true` says the two differ — raise `limit` to see the rest. A project with 300 tests answers `count: 300, returned: 100, truncated: true` under the default. Never derive the suite size from `len(tests)`.
>
> ⚠️ **`count` means the opposite thing in `test_list`.** There `count` is the *returned* length and `total` is the discovered size; here `count` is the *discovered* size and `returned` is the returned leng

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.