> ## Documentation Index > Fetch the complete documentation index at: https://docs.noqa.ai/llms.txt > Use this file to discover all available pages before exploring further. # Full-cycle testing > Say what to cover and your coding agent does the testing: it learns the app, writes the cases and runs them until they pass. ## What you do [Install the CLI and the `noqa-testing` skill](/docs/cli#setup), then ask. The skill carries the whole workflow, so there is nothing else to set up and nothing to explain to your agent: ``` Cover the main flows of my app with tests ``` ``` Write a test for the paywall: open it from Settings, subscribe monthly, then restore purchases ``` Both work — the first leaves the scope to the agent, the second pins it down. From there you only answer when it asks. ## What the agent does * **Learns your app** from three sources: * your **code** — the screens, navigation and business rules behind the feature; * your **tests in noqa** — the app context, the cases you already have, and their earlier runs, which show the screens and the paths between them as your app was really used; * the **app itself** — where the first two leave gaps, it builds the app, installs it and walks through it on a device. * **Writes the test cases** — in plain English, from a cold launch on a clean install, with shared steps like login and onboarding pulled into reusable flows. * **Proves each case before saving it** — it performs the scenario on the device first, so the steps match what the app really does, not what the code suggested. * **Runs the suite and fixes what it wrote** — the noqa agent executes the cases; a case that fails on its own wording comes back for a fix and another run. * **Reports real defects** — a failure in the app itself is handed to you with the failing step and its screenshot, and the case is left as written. What you get back is a suite where every case has been walked once and verified once — reliable from the first release, not the second. ## Regression before release Once the cases exist, running them is one command. The noqa agent runs one case or the whole suite against a fresh build — on your local devices, on cloud devices, or in CI, so every release passes the same checks. How the agent drives a device What a good case looks like