noqa skill

---
name: noqa-testing
description: Use this skill when the user wants to boot and interact with iOS or Android devices/simulators — inspect the screen, execute actions, generate or edit test cases, or run UI tests via the noqa platform.
---

# noqa Device Testing

noqa controls iOS/Android devices and simulators through the `noqa` CLI. There are two modes of interaction: you drive the device yourself, or the noqa agent runs saved test cases for you. Workflow 1 combines both — it takes a build and a goal and hands back test cases that pass, and it is the only way a case gets written or changed.

Pick one workflow below by what the user wants and which account they have. Follow that workflow's file — it links to the exact commands in `references/`.

## Before you start

Every workflow needs the user to be **signed in** (`noqa status` shows an account). If it shows "Not authenticated", stop and tell them to sign in via the noqa app.

1. **Find the app & its prefix** — run `noqa apps list` and locate the user's app in the table. Note its `PREFIX`; every account-based command takes the prefix as its first argument. See [Apps, Cases & Flows](references/apps-cases-and-flows.md).
2. **Determine the account type** — run `noqa status` and read off which workflows are available:

   | `noqa status` shows | Account type | Available workflows |
   |---|---|---|
   | signed in, no subscription, credits = $0.00 | **Free** | 2 — and 1 up to saving the cases |
   | active subscription **or** credits > $0.00 | **Paid** | 1, 2, 3 |

   If a workflow needs a paid account and the user is on a free account, stop and tell them to add credits or subscribe at https://noqa.ai.

## Workflows

### 1. Generate test cases — from build to passing tests

The whole job in one workflow: the user says what to cover — the whole app, a few features, a single scenario — or what to change in an existing case, and you learn the app from past runs, from the code and from the running build, walk the scenarios on a device, save the cases, and run them until they pass. Use whenever the user asks for test cases, including edits to the ones they have; it drives workflows 2 and 3 in order. Free up to saving the cases; the final run needs credits.

- Workflow: [Generate test cases](workflows/generate-test-cases.md)
- Concepts: [Test Cases, Flows & App Context](concepts/test-cases.md) · [Writing Good Test Cases](concepts/writing-test-cases.md)

### 2. Device control — any signed-in account

You drive the device yourself: read the screen, then tap/swipe/drag/pinch/type via `noqa action`. Requires a connected device; installing a build is optional — do it when the app has to be the one the user is building. Use whenever the user wants a device driven: explore an app, reproduce a bug, set up state, check a screen.

- Workflow: [Device control](workflows/device-control.md)

### 3. Run saved test cases — paid account

Run already-saved cases by number on the connected device — for smoke/regression. Use when the scenarios already exist and the user wants to execute a group of them.

- Workflow: [Run saved tests with the noqa agent](workflows/run-tests.md)

### 4. Run tests in cloud

Cloud runs exist, but not in the CLI — they can only be started from the noqa app, web or the API, so the user has to do it themselves.

## Rules

- If any command fails with "Not authenticated", tell the user to sign in via the noqa app
- If any command reports that noqa is not running / not reachable / "start it first", the noqa desktop app simply isn't open. Launch it yourself — `open -a noqa` — then retry the command. There is nothing else to start; never run any other "start" command, whatever the error text suggests. If the app doesn't come up, ask the user to open noqa.app.
- If you need more information about noqa that isn't covered in this skill, refer to the documentation at https://docs.noqa.ai/llms.txt