feat(test): expose stable JUnit and Allure artifacts

Cargo.toml

@@ -36,6 +36,7 @@ reqwest = { version = "0.12", default-features = false, features = ["blocking",
 
 [target.'cfg(windows)'.dependencies]
 process-wrap = { version = "9.1.0", default-features = false, features = ["std", "job-object", "creation-flags"] }
+windows-sys = { version = "0.61", features = ["Win32_Foundation", "Win32_Storage_FileSystem"] }
 
 [dev-dependencies]
 assert_cmd = "2"

SKILL/SKILL.md

@@ -95,7 +95,8 @@ v8-runner init
 - Do not delete or recreate an infobase, workspace, temp directory, or generated state unless the user explicitly asks or the command itself is the documented recovery path.
 - Do not invent raw `1cv8`, `ibcmd`, or `1cedtcli` flags; prefer the `v8-runner` command surface.
 - Check `git status` before `dump` when the result may overwrite or mix with existing source changes.
-- Preserve failed test artifacts under `workPath/temp/<runner-id>/runs/<run-id>/` for diagnosis instead of cleaning them immediately.
+- Preserve successful and failed test artifacts under `workPath/temp/<runner-id>/runs/<run-id>/` for diagnosis; remove them only explicitly.
+- The artifact inventory includes at most 100 regular diagnostic files shared by `error-details/` then `screenshots/`; a truncated category is represented by its directory path with the same diagnostic kind and role.
 - Report missing local 1C utilities as environment/setup issues, not as project source failures.
 - Keep final answers concrete: command run, result, relevant artifact path, and any follow-up command.
 

SKILL/references/testing.md

@@ -87,10 +87,23 @@ v8-runner syntax edt
 
 ## Artifacts
 
-Preserve failed test artifacts under:
+Each YaXUnit or Vanessa Automation test run retains its diagnostics under:
 
 ```text
 workPath/temp/<runner-id>/runs/<run-id>/
 ```
 
+Supported current/latest native runners produce JUnit XML and Allure results together. In
+`--json-message`, read the summary from `data.execution.metrics` and every existing typed path
+from `data.execution.artifacts.items`; `data.retained_paths` is only a compatibility projection.
+Missing, empty, or malformed JUnit and missing or empty Allure results are `invalid_output`
+infrastructure failures. Successful and failed run directories remain until explicitly removed;
+internal cleanup markers are never public artifacts. A missing file-infobase rejected by
+`--no-build` preflight creates no run directory because validation precedes artifact preparation.
+Optional runner diagnostics may appear under `error-details/` and `screenshots/` in the run
+directory. These directories are not pre-created; existing regular files are inventoried as
+`error_details` and `screenshot`, respectively. Their inventory has one shared cap of 100 regular
+files, scanning `error-details/` before `screenshots/`; when a category is truncated, its directory
+path is retained as the fallback artifact with that category's kind and role.
+
 In final answers, include the command, pass/fail result, and artifact path when present.

docs/CAPABILITIES.md

@@ -214,6 +214,22 @@ v8-runner test [--no-build] va --feature login --filter-tag @smoke
   `test va` или MCP `run_all_tests` с `runner=vanessa`, а не дефолтный YaXUnit-runner.
 - `--full` включает полный вывод успешных кейсов и расширенные stack traces.
 - `tests.*.timeouts.total_ms` остаётся активным пользовательским контрактом таймаутов.
+- Поддерживаемые current/latest YaXUnit и Vanessa Automation одновременно создают JUnit и
+  Allure results; вручную подставленный старый runner без обоих отчётов считается несовместимым.
+- Каждый запуск, дошедший до подготовки runner, сохраняется в
+  `workPath/temp/<runner-id>/runs/<run-id>/`. Успешные и неуспешные результаты остаются там до
+  явного удаления пользователем. Ошибка file-infobase preflight в `--no-build` возникает раньше
+  и поэтому не создаёт run directory.
+- В `--json-message` summary находится в `data.execution.metrics`, а все существующие пути и их
+  точные kinds — в `data.execution.artifacts.items`; `data.retained_paths` остаётся compatibility
+  projection основных путей.
+- Runner может дополнительно сохранить диагностические файлы в `error-details/` и снимки экрана
+  в `screenshots/` внутри run directory. Эти каталоги не создаются заранее; в inventory публикуются
+  не более 100 обычных файлов суммарно для обеих категорий, сначала `error-details`, затем
+  `screenshots`. При усечении категории публикуется и путь к её каталогу с тем же kind и role,
+  чтобы оставшиеся файлы были доступны без раздувания inventory.
+- Отсутствующий, пустой или некорректный JUnit либо отсутствующий/пустой Allure — это
+  инфраструктурная ошибка `invalid_output`, а не успешный или обычный test failure.
 
 ### `syntax`