01 / CONNECT
Install the skill.
Run zeuz skills install to add the skill to supported agents detected on your machine. It lives in their user skill directories, ready for ZeuZ testing tasks across projects.
Test automation, from your terminal
Your agent.ZeuZ testing.Together.
Create tests, run suites, and inspect results from the AI tools you already use. ZeuZ CLI brings your agent into your testing workflow.
ONE CLI. YOUR EXISTING TOOLS.EXAMPLE WORKFLOW
>_ zeuz
$ zeuz skills list $ zeuz skills install
Find supported agents on your machine and install the ZeuZ skill in their user directories.
Commands shown are examples. Tests run in your configured ZeuZ workspace.
Works alongside your coding tools
- Claude Code
- Codex
- Gemini CLI
- Copilot CLI
- Cursor
- opencode
A session with Claude Code
The feature changed.
Now test the behavior.
A developer finishes a sign-in update and asks Claude Code for test coverage. Follow the handoff to ZeuZ, from the first request to a report the team can review.
Claude Code~/projects/storefront
Skill · zeuzReady to play
01 / 05
Read the full example transcript
Developer We just updated sign-in. Use ZeuZ to add smoke coverage in staging: a valid user reaches the dashboard and an invalid password shows an error. Reuse our managed variables, run the tests, and show me the results. Claude I'll check the ZeuZ workspace and existing coverage, then create and run the missing tests. Claude → Skill(zeuz) ZeuZ testing instructions and CLI references loaded. Claude → Bash $ zeuz workspace show --workspace staging --json Output summary: Workspace: staging Test root: ~/.zeuz/agent/clones/<workspace-id> Claude → Bash $ zeuz list tests/auth --workspace staging --json Output summary: Existing coverage: password reset No sign-in smoke cases in this example workspace. Claude → Bash $ zeuz vars list --workspace staging --json Output summary: Managed inputs: app_url, qa_username, qa_password Sensitive values are masked. Claude The workspace has the inputs we need. I'll ask ZeuZ to save both sign-in cases, then run them from the testing workspace. Claude → Bash $ zeuz --workspace staging "Create two browser smoke cases for sign-in. Use managed app_url, qa_username, and qa_password. Verify that valid credentials reach the dashboard and an invalid password shows an error. Reuse shared steps and register variable usage. Save the tests; do not run them." Output summary: Saved: tests/auth/sign_in/test_sign_in.py Cases: valid sign-in, invalid password Reused shared steps: open_app, sign_in Managed-variable usage registered. Claude ZeuZ saved the two cases. I'll run the test file and inspect both the test results and report upload. Claude → Bash $ zeuz run tests/auth/sign_in/test_sign_in.py --workspace staging --headless --json Output summary: PASS Valid user reaches the dashboard PASS Invalid password shows an error Report upload: complete Report ID: <report-id> Claude Both cases passed. The report upload completed; I will read the recorded evidence next. Claude → Bash $ zeuz reports show <report-id> --workspace staging --json Output summary: Cases: 2 passed Assertions: dashboard reached; invalid password rejected Recorded evidence: step results, screenshots, browser trace Claude Sign-in smoke coverage is ready. ✓ Valid credentials reached the dashboard. ✓ An invalid password displayed the expected error. The two cases are saved in ZeuZ. Your team can review the published report and its recorded artifacts in Test Studio.
An illustrative session using supported ZeuZ commands. Responses and results are scripted and condensed; this page does not execute commands. ZeuZ access, model access, the skill, and staging resources are already configured in this example.
The workflow
From a request
to a test result.
The ZeuZ skill gives your agent practical guidance for using the CLI: how to find existing tests, author new coverage, use shared resources, and interpret results. Describe the behavior and expected outcome in the conversation where you already work.
02 / TEST
Ask for the behavior.
Choose the workspace, application, and expected outcomes. Your agent can use direct CLI commands or delegate test authoring to ZeuZ's agent, then run the selected cases or suite.
03 / REVIEW
Read the evidence.
Inspect case results, failed steps, and available screenshots, traces, and logs. Use published Test Studio reports to bring the same evidence into your team's review.
ZeuZ Test Studio
Your team's testing,
in one workspace.
Give agents access to the testing context your team already maintains. Tests, environment values, input files, and run evidence stay part of a shared ZeuZ workflow, whether the work starts in a terminal or Test Studio.
- 01 / COVERAGETests & reusable steps
- List existing cases, author browser and API tests, and combine both in an end-to-end flow. Reuse shared steps for common actions such as signing in or preparing test data.
- 02 / CONFIGURATIONManaged variables
- Read and maintain named workspace values, organize labeled options, and select the value a test uses. Keep application URLs and environment configuration reusable across cases.
- 03 / TEST DATAAttachments & fixtures
- Use shared input files for upload flows, imports, and data-driven scenarios. Agents can inspect registered attachments and add the fixtures needed for a test.
- 04 / EXECUTIONTest sets & schedules
- Group cases into ordered sets, run a focused smoke suite, or manage recurring automations with a defined schedule or webhook trigger. Keep the target and cadence explicit.
- 05 / EVIDENCEReports & run history
- Inspect finished reports and active runs, follow execution logs, and download recorded artifacts. Review failures against the test's assertions and the evidence collected during execution.
- 06 / CONTEXTProduct knowledge
- Search indexed workspace knowledge for features, endpoints, and test strategy. Bring registered product repositories and documentation into test planning through ZeuZ's ingestion workflow.
Put it to work
Keep testing close
to the work.
Use ZeuZ during feature development, when organizing regression coverage, or while investigating a failing run. Your agent can carry the testing task forward with a concrete request.
01 / DEVELOPAfter a feature change
Turn a change into lasting coverage.
Describe the new behavior and the result users should see. Ask your agent to check existing coverage, create the missing test, and verify it against the intended environment.
Try asking your agent“Use ZeuZ to test our updated password reset flow in staging. Verify that an expired link is rejected and a valid link lets the user set a new password.”
02 / ORGANIZEBefore a release
Build a suite your team can reuse.
Find relevant cases and arrange them into a named test set. Reuse managed variables and shared fixtures so the same suite has a clear, repeatable setup.
Try asking your agent“Find our checkout tests in ZeuZ. Put the happy path and payment validation cases into a Checkout smoke set, then run it against staging.”
03 / INVESTIGATEWhen a run fails
Bring the evidence into the conversation.
Ask for the failed steps, available artifacts, and a grounded explanation. Review a proposed test repair against the original expected behavior before applying it.
Try asking your agent“Inspect the latest Checkout smoke report. Summarize the failed steps, link the run evidence, and explain what to investigate before changing the test.”
These are example requests. Use your actual application behavior, workspace, and test names; your agent follows the permissions and approvals set by its host.
Built for the terminal, too
Use a prompt.
Or a command.
The skill helps agents choose the right operation. The CLI also works directly: discover tests, run a set, and inspect results with commands you can use in scripts and build pipelines.
Give ZeuZ a focused task.
A quoted prompt runs through ZeuZ's agent and exits when the task finishes. Your external agent can delegate a testing request this way while continuing to handle the surrounding development work.
For a conversation in the terminal, launch zeuz without a prompt. For routine operations, use the dedicated subcommands and request structured output with --json.
Use --workspace for a specific target. With a saved selection, the CLI resolves your default workspace for normal testing commands.
Example · A single delegated task
$ zeuz --workspace staging "Create a login smoke test using our managed variables. Run it and summarize the results."Include the behavior, expected assertions, and environment in your request. Delegated prompts use your configured model access.
Find existing coverage
zeuz list --workspace staging --jsonDiscover cases in the selected testing workspace before adding new coverage.
Run a focused suite
zeuz run "Checkout smoke" --workspace staging --headless --jsonExecute a named test set and collect structured results for the next step in your workflow.
Inspect a report
zeuz reports show <report-id> --workspace staging --jsonRead recorded results using an ID returned by zeuz reports list.
See shared test inputs
zeuz vars list --workspace staging --jsonInspect managed variable names and options. Sensitive values are masked in normal listing.
Replace example names and IDs with your own. Use zeuz --help or a subcommand's --help for the options in your installed version.
Start with the tools you have
Add ZeuZ to
your agent's toolkit.
One command installs the portable skill for supported agents detected on your machine. You can choose individual tools and inspect the planned changes first.
- Check your CLI. Run
zeuz --helpto confirm ZeuZ is installed.zeuz doctorhelps check your local environment and saved configuration. - Install and discover.Add the skill, then start a fresh agent session if it has not appeared in the host's skill list.
- Choose your testing context. Configure your ZeuZ access and workspace, then ask for a focused test task. Skill installation itself needs no ZeuZ login.
zeuz skills installSupported: Claude Code, Codex, Gemini CLI, Copilot CLI, Cursor, and opencode. Installed into the host's supported user skill location.
Updates protect local edits. An explicit forced replacement retains a backup; your agent's own skill settings and permissions still apply.
Before your first task
A few useful details.
The CLI connects your existing agent workflow with ZeuZ's testing capabilities. Here is what to expect from installation, execution, and updates.
What does the skill install?
A portable set of ZeuZ instructions and references in your supported agent's skill directory. It covers test authoring, execution, workspace resources, and result inspection. The ZeuZ CLI must already be installed and accessible to your agent.
Can I keep using my current coding agent?
Yes. Claude Code, Codex, Gemini CLI, GitHub Copilot CLI, Cursor, and OpenCode can use the skill. Your agent can call ZeuZ subcommands directly or delegate a testing task through a ZeuZ prompt.
Will my agent pick up the skill automatically?
The installed skill describes when it is relevant to ZeuZ testing work. Your host decides when to load it, subject to its own discovery rules, disabled skills, and consent settings. Start a fresh session if a newly installed skill is not visible.
Do I need to select a workspace every time?
Normal testing commands use the saved CLI selection or your Test Studio default when available. Pass --workspace <name-or-id> to target a particular workspace for one operation. Skill installation does not need a testing workspace.
Can I use it in a script or build pipeline?
Yes. Direct commands support operations such as listing cases, running tests, and reading reports, with --json where available. Configure access and the workspace ahead of time. A supplied prompt runs once and exits; delegated prompts also need configured model access.
What kinds of tests can I work with?
Author and run browser tests, API tests, or flows that combine both. Cases can reuse shared steps, managed variables, and registered input attachments. Existing test sets help organize focused smoke coverage and broader regression runs.
How do updates work?
zeuz skills update refreshes installed copies from your current CLI version. Upgrade the CLI for newer skill content. Edited copies require explicit replacement, which keeps a backup. Existing linked installations remain managed by their original source.