AI agent skill (test-lab-plan)

test-lab-plan is an agent skill that teaches Claude Code, Codex, or Cursor how to turn a described flow into a production-ready test-lab plan: an explicit URL, declarative acceptance criteria, {{credentials.<key>}} for secrets, and {{data.<fixture>.<field>}} for generated data.

Install

First install the testlab CLI, which provides the testlab command:

npm i -g @test-lab-ai/cli

Then add the skill to your agent of choice:

testlab skills install                  # auto-detects the agent(s) you use
testlab skills install --agent codex    # Codex -> .agents/skills
testlab skills install --agent cursor   # Cursor -> .cursor/rules/test-lab-plan.mdc
testlab skills install --agent all      # all three

With no --agent, it detects which agents you use (the agent running the command, plus the .claude / .agents / .cursor folders in your project) and installs for each, defaulting to Claude Code if none are found.

Add --global to install under your home directory instead of the project (Claude Code and Codex only; Cursor user rules are configured in Cursor's Settings). Restart your agent to load it. The skill is bundled with the CLI and version-locked to it, so installs are offline and reproducible. The public Test-Lab-ai/skills repo is the browsable, open-source source.

Keeping it current: a skill change ships as a new CLI version, so npm i -g @test-lab-ai/cli@latest updates it. The CLI auto-refreshes installed copies the first time you run it after an upgrade, and testlab skills update re-installs the bundled version on demand.

How it works

When you ask your agent to "write a test-lab test for X", the skill:

1. Reads the real source when the target site is in your repo, so the acceptance criteria quote actual on-screen text and response shapes instead of guesses.
2. Drafts one plan with a single label, the right mode (Quick or Deep), and the credential and data references it needs.
3. Offers to create it directly in your account with the testlab CLI, or hands back a paste-ready plan for the dashboard.

Two ways to use it

Author a test while building a feature. You just added a page or route, so you ask the agent to write a test for it. The skill reads the new component, produces a grounded plan, and creates it through the CLI.

Import the tests you already have. Point the agent at Playwright or Cypress specs, Cucumber features, a TestRail or Zephyr export, or a prose doc. It converts each into a plan and imports the whole batch with testlab import.

Example prompts

You invoke the skill and describe the test in plain language; it writes the plan JSON and runs the CLI. In Claude Code that means /test-lab-plan; in Codex or Cursor, mention test-lab and the skill triggers.

Create a test for a feature you just built:

/test-lab-plan write a test for the signup flow at https://app.example.com/signup, then create it in my account

Import existing tests:

/test-lab-plan convert the Playwright specs in ./tests/e2e into test-lab plans and import them

The skill surveys your account first, so it reuses the matching project, existing credentials, and existing fixtures instead of duplicating them.