# critic.selftest

> Talos next · Since next

Play scripted scenarios headless (argus://docs/agent#self-test-scenarios): press, hold actions, wait for a screen, text or goals.

Per step pass/fail and what it read; the failing step, why, frame reached. No scenario: every selftest/*.json. Reads files: save first.

| | |
|---|---|
| Title | Self-test flows |
| Mode | `disk` |
| Token cost | about 172 tokens (listed only when its toolset is enabled) |
| Toolset | `selftest` |

## Arguments

| Argument | Type | Required | Default | Allowed values | Description |
|---|---|---|---|---|---|
| `scenario` |  | no |  |  | A selftest/ name, a project-relative .json, or the scenario |
| `scene` | `string` | no |  |  | The startup scene when absent |
| `timeout` | `integer` | no | `60` |  | Wall seconds per scenario |

## Example

An MCP `tools/call` with the required arguments (placeholders in angle brackets).

```json
{
  "name": "critic.selftest",
  "arguments": {}
}
```