Skip to content

QA Testing

Spec-driven testing with structured reports or production-ready Playwright code.

Best for: Pre-release testing, E2E automation
Time estimate: 15-45 minutes
Skills used: meow:qa-manual, meow:qa, meow:browse
Browser skills: meow:agent-browser, meow:playwright-cli

Overview

MeowKit offers two QA approaches: meow:qa (find bugs + fix them + verify fixes) and meow:qa-manual (spec-driven testing with reports or E2E code generation). Both use browser automation under the hood.

Manual QA report

Step 1: Run from spec

/meow:qa-manual tasks/plans/260327-auth.md --report

Step 2: Agent navigates like a human

The skill routes each action to the best browser tool:

Step 3: Auth handling

When the agent encounters a login page:

⚠️ Authentication required at: https://app.example.com/login
meow:qa-manual needs credentials to continue.
Please provide:
- Email: ___________
- Password: ___________
Type 'skip' to skip this flow or 'abort' to stop testing.

The agent never guesses credentials. You provide them, it fills the form and continues.

Step 4: Review the report

markdown
## QA Manual Test Report
Spec: tasks/plans/260327-auth.md
Tested at: 2026-03-27 14:30 UTC

| Flow | Steps | Passed | Failed | Status |
|------|-------|--------|--------|--------|
| Login | 5 | 5 | 0 | PASS |
| Password reset | 4 | 3 | 1 | FAIL |

### Failed: Password reset — Step 3
Expected: "Reset email sent" message
Actual: 500 Internal Server Error
Evidence: /tmp/password-reset-error.png

E2E code generation

/meow:qa-manual tasks/plans/260327-checkout.md --generate

Same navigation, but outputs .spec.ts files with role-based locators (getByTestId, getByRole, getByLabel).

Systematic QA (find + fix)

/meow:qa https://app.example.com

The meow:qa skill goes further: it finds bugs, fixes them in source code, commits each fix atomically, re-verifies, and produces a before/after health score.

Next workflow

Documentation — keep docs in sync

Released under the MIT License.