# CI test scripts

> How do I set up CI tests in Windmill? Write test scripts that automatically run when the tested script or flow is deployed.

Add a `test:` annotation at the top of any script to turn it into a CI test. When the tested [script](../../getting_started/0_scripts_quickstart/index.mdx) or [flow](../../getting_started/6_flows_quickstart/index.mdx) is deployed, the test runs automatically.

## Writing a test script

### Single target

```typescript
// test: script/u/admin/my_script

  if (result !== 42) {
    throw new Error(`Expected 42, got ${JSON.stringify(result)}`);
  }

  return result;
}
```

### Multiple targets

```typescript
// test:
// script/u/admin/script_a
// script/u/admin/script_b
// flow/u/admin/my_flow

export async function main() {
  // test all items
}
```

### Wildcards

Annotations support glob patterns so a single test script can cover a whole folder tree:

- `*` matches one path segment.
- `**` matches any depth.

```typescript
// test: script/u/admin/*
// test: flow/u/team/**
```

The first pattern matches every script directly under `u/admin/`. The second matches every flow at any depth under `u/team/`. Wildcards can be combined with multi-target annotations on separate lines.

### Branching on the triggering runnable

CI test jobs receive a `WM_TESTED_RUNNABLE` environment variable containing `{kind}/{path}` of the runnable that triggered the test (for example, `script/u/admin/my_script`). This lets a single test script cover multiple targets and branch on which one fired it.

```typescript
// test: script/u/admin/*

export async function main() {
  const tested = process.env.WM_TESTED_RUNNABLE!; // e.g. "script/u/admin/my_script"
  const path = tested.replace(/^script\//, "");
  return await wmill.runScript(path, {});
}
```

### Python

```python
# test: script/u/admin/my_script
import wmill

def main():
    result = wmill.run_script("u/admin/my_script", {})
    assert result == 42
    return result
```

In Python, the triggering runnable is exposed as `os.environ["WM_TESTED_RUNNABLE"]`.

The annotation uses the comment syntax of each language (`//` for TypeScript, `#` for Python, etc.).

The script creation page also provides ready-made CI test templates for TypeScript and Python.

## Where results appear

Test scripts show a yellow **CI test** badge in the script list and detail page.

On the detail page of the tested script or flow, each test is listed with its status (pass, fail, or running) and a link to the job run. Results auto-refresh while tests are running.

In [workspace forks](../20_workspace_forks/index.mdx), CI results are also visible:

- The fork banner shows a summary of passing, failing, and running tests across all changed items.
- On the [comparison page](../20_workspace_forks/index.mdx#merge-workspaces-from-the-ui-merge-ui), each changed item displays a CI badge, and a test summary lists all test scripts with their latest results.