Tests features end-to-end via browser QA, API calls, and LLM eval tests.
npx skills add https://github.com/elie222/inbox-zero --skill test-feature
Please help me install this Agent Skill. Skill: test-feature Source: https://github.com/elie222/inbox-zero/tree/main/.claude/skills/test-feature Command: npx skills add https://github.com/elie222/inbox-zero --skill test-feature
Verifying that a newly built feature actually works requires juggling browser testing, API verification, test data setup, and AI output evaluation, which is slow and error-prone when done manually.
__tests__/eval/ using describeEvalMatrix and judgeMultiple to assess AI output quality, and can persist tests as reusable QA flows.Ask the assistant to test the feature you just built, for example by saying "test the new bulk archive flow end to end and show me screenshots."
High-intent search queries and answers about installing and using this skill.
Describe the feature or point to recent code changes, and the skill plans an approach combining browser QA, API calls, or eval tests. It sets up the dev server, creates test data, executes the test, and reports pass/fail with screenshots.
Use headless Chrome launched with a saved user profile directory that already has a logged-in session, then connect agent-browser over CDP on port 9222. Google OAuth blocks the built-in Chromium, so real Chrome with an existing session is required.
Yes, the gws CLI creates test data such as Drive folders and files, Gmail messages, and Calendar events from the command line. The browser is then used only for app-specific configuration that exists in the product UI.
Eval tests live in __tests__/eval/ and use describeEvalMatrix for cross-model comparison plus judgeMultiple with criteria like ACCURACY, TONE, and NO_HALLUCINATION. Run them with pnpm test-ai followed by the eval test name.
NEXT_PUBLIC_* environment variables are baked in at build time, so adding one to .env while the dev server runs has no effect. You must restart the dev server before testing for the flag to apply.
Persist tests that verify important behavior someone might break later, either as a browser flow spec in qa/browser-flows/ or an eval test in __tests__/eval/. Trivial one-off checks like page loads should not be persisted.
Andresch2Create isolated Git worktree workspaces with dependency installation and baseline tests.
twentyhqTests pull requests in a real browser and verifies database and log effects.
crbnosCapture browser errors with screenshots and element snapshots during end-to-end testing.
diegocamara89Enforce test-driven development with unit, integration, and E2E tests.