AgenTeX (Agentic Test eXecution) takes manual test execution off your plate. Instead of clicking the same scenarios by hand, an agent plans them, drives a via @playwright/cli or a via Appium,...
概览
AgenTeX (Agentic Test eXecution) takes manual test execution off your plate. Instead of clicking the same scenarios by hand, an agent plans them, drives a via @playwright/cli or a via Appium, captures screenshot/log evidence, and produces a consolidated defect report — either (human-in-the-loop) or in (autonomous, one session per spec file). It . New here? walks you through install → browser driver → /init-test → permissions → first run. The short version: See docs/ for the full reference on any feature. 11 commands, 17 skills and 3 subagents. Solid arrows: a command runs a skill. Dotted arrows: work handed to a subagent. An arrow means the skill reads the other skill's instructions or runs its scripts. Arrows both ways mark two skills that depend on each other. Every run writes to a timestamped executions/execu_ / folder — report.md, extent-report.html, per-session logs/screenshots, and a merged bug list. New to the codebase?
README
以下为原文 README 节选，完整内容请查看原文。
AgenTeX
Agentic QA for Claude Code — an agent plans, runs, and reports your tests so you don’t click through them by hand.
AgenTeX (Agentic Test eXecution) takes manual test execution off your plate. Instead of clicking the
same scenarios by hand, an agent plans them, drives a real browser via
@playwright/cli or a real native mobile
app via Appium, captures screenshot/log evidence,
and produces a consolidated defect report — either sequentially (human-in-the-loop) or in
parallel (autonomous, one session per spec file). It never modifies your application code.
Getting Started
New here? Getting Started walks you through install → browser driver
→ /init-test → permissions → first run. The short version:
/plugin marketplace add MhmdElGazzar/elgazzar-plugins
/plugin install agentex@elgazzar-plugins
/init-test
/execute-test https://example.com
Features — how each one works
| Feature | How it works | Docs |
|---|---|---|
| Test execution | An agent plans scenarios and runs each step through its driver — a real playwright-cli browser, your APIs, or your database (API/DB-only specs need no browser) — capturing evidence and reporting defects, sequential (approve each step) or parallel (one qa-executor subagent per spec file). |
test-execution |
| Define flow | /define-flow builds a spec by doing it: an agent-led session proposes each step, executes it live the moment you agree, and you assert the real outcome before the next step — the saved spec follows the normal conventions and runs unmodified via /execute-test. Point it at an existing spec to walk it through and clarify it. |
define-flow |
| Mobile testing | Same flow for native Android/iOS apps, driven through a real Appium session (raw WebDriver REST or a bundled webdriverio wrapper) — sequential or parallel (one mobile-qa-executor subagent per spec file, per device/emulator). |
mobile-testing |
推荐工具
换一个关键词，或者移除筛选条件。
安装
npx skillfish add mhmdelgazzar/agentex