---
title: "Agent-Driven Testing — Overview"
description: "Shiplight is a single /shiplight skill your coding agent runs, backed by an MCP server that gives it a real browser. Describe what you want in plain language and the agent picks the right workflow: verifying UI, authoring tests, and reviewing your app."
---

# Agent-Driven Testing

<div class="view-markdown-wrapper">
<ViewMarkdown />
</div>

Shiplight ships as a single **`/shiplight` skill** for your coding agent, backed by the [Shiplight MCP server](/local/agent-workflow/tool-reference) that gives it a real browser to drive. Ask it to verify a UI change on the spot, or set up auth and create YAML tests. From there, [Local Testing](/local/run-locally) takes over for running, fixing, and CI. It can also run a broader quality review whenever you need one.

As you work together, Shiplight preserves durable facts learned from your conversations as [project knowledge](/local/agent-workflow/project-knowledge). Future workflows read these project-local notes so you do not have to repeatedly explain the same application quirks, environment constraints, or testing preferences.

## Invoking Shiplight

Run `/shiplight <subcommand> [context]`, or just describe what you want in plain language. The agent picks the right workflow:

```
/shiplight create-yaml-tests for the sign-up flow at http://localhost:3000
/shiplight screenshot the dark mode change    # natural phrasing resolves to `verify`
```

Run `/shiplight` alone for the full menu.

::: tip Prerequisite
Install the skill and MCP server once per agent. See [Quick Start](/getting-started/quick-start).
:::

### Subcommands

| Group        | Subcommand                  | What the agent does                                                                                                | Guide                                                  |
| ------------ | --------------------------- | ------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------ |
| **Setup**    | `init`                      | Scaffold a Shiplight test project and write `specs/context.md`                                                     | —                                                      |
|              | `auth`                      | Set up or repair login and saved storage state                                                                     | [Authentication](/local/agent-workflow/authentication) |
|              | `update`                    | Refresh installed Shiplight skills                                                                                 | —                                                      |
| **Author**   | `create-yaml-tests`         | Write YAML E2E tests by walking your app in a real browser                                                         | [Create Tests](/local/agent-workflow/create-tests)     |
|              | `create-agent-verification` | Author an agent verification — drives a full user journey, then proves the backend state behind it (API, DB, logs) | [Agent Verification](/local/agent-verification)        |
|              | `cover`                     | Decide what to test and the strategy, drive the producers, and report coverage                                     | [Decide what to test](/guides/decide-what-to-test)     |
| **Maintain** | `fix`                       | Reproduce failing or drifted tests, diagnose root cause, repair the YAML, report app bugs                          | [Fix a failing test](/guides/fix-a-failing-test)       |
| **Check**    | `verify`                    | Visually confirm UI changes in the browser during local development                                                | [Verify a change](/guides/verify-a-change)             |
| **Review**   | `review`                    | App-quality review across security, privacy, compliance, design, resilience, performance, SEO, GEO                 | [Review your app](/guides/review-your-app)             |
| **Ship**     | `ci`                        | Wire CI workflows and the failure-triage pipeline                                                                  | [CI/CD](/local/ci)                                     |
|              | `cloud`                     | Read Shiplight Cloud (Nova) results: runs, failing/flaky tests, artifacts                                          | —                                                      |
| **Help**     | `help`                      | List or explain subcommands. Never executes                                                                        | —                                                      |

The main workflow for most teams is **authoring tests** ([Create Tests](/local/agent-workflow/create-tests)); the rest support it. Task-by-task walkthroughs — verifying changes, mobile emulation, persistent profiles, extensions, attaching to a browser — live in the [Guides](/guides/) and apply across all of the above; underlying tool parameters live in [MCP Tool Reference](/local/agent-workflow/tool-reference).