██████╗ █████╗ ████████╗ ██████╗ ██████╗ ██╗ ██╗ ██╗██╗████████╗ ██╔═══██╗██╔══██╗ ╚══██╔══╝██╔═══██╗██╔═══██╗██║ ██║ ██╔╝██║╚══██╔══╝ ██║ ██║███████║█████╗ ██║ ██║ ██║██║ ██║██║ █████╔╝ ██║ ██║ ██║▄▄ ██║██╔══██║╚════╝ ██║ ██║ ██║██║ ██║██║ ██╔═██╗ ██║ ██║ ╚██████╔╝██║ ██║ ██║ ╚██████╔╝╚██████╔╝███████╗██║ ██╗██║ ██║ ╚══▀▀═╝ ╚═╝ ╚═╝ ╚═╝ ╚═════╝ ╚═════╝ ╚══════╝╚═╝ ╚═╝╚═╝ ╚═╝
AI-powered QA that analyzes your codebase with parallel agents and produces a browsable dashboard — findings, screenshots, videos, fix lists and Jira-ready tickets.
$ npx qa-toolkit-ai init
Install the skills into your project's .cursor/, say one word, and watch
three agents sweep the codebase in parallel.
you: /qa ▸ qa-frontend scanning components, a11y, responsive… done ▸ qa-backend scanning API, validation, errors… done ▸ qa-e2e driving the site, capturing video… done QA REPORT ───────── findings: 3 critical · 11 major · 24 minor evidence: 18 screenshots · 4 E2E videos split: functional (PM) vs technical (devs) ✓ dashboard docs/qa-dashboard (Docsify) ✓ fix list prioritized, effort-estimated ✓ tickets Jira-ready markdown ✓ share public link via cloudflare tunnel
Frontend, backend and E2E agents run at the same time, each with its own checklist and report format.
Every finding lands in a Docsify site — filterable, organized, and easy to scroll in a review meeting.
Findings split into a PM/client view and a developer view. Nobody reads what they don't need.
Screenshots at desktop, tablet and mobile widths, plus E2E videos of actual broken flows.
Generates Jira-ready tickets and Confluence checklists from the findings. Paste and assign.
A Cloudflare tunnel exposes the dashboard publicly, so stakeholders without repo access can browse it.
Run everything, or aim at one layer.
| Command | What it runs |
|---|---|
/qa | Full sweep — all three agents + merged report |
/qa-frontend | Components, accessibility, responsive behavior |
/qa-backend | API contracts, validation, error handling |
/qa-e2e | Real user flows, captured on video |
npx qa-toolkit-ai init | Install skills, rules and dashboard template into .cursor/ |
The decisions worth knowing about, written down.
An agent that says "looks mostly fine" in a chat window is useless a week later. The dashboard makes findings a durable artifact.
Every finding ships with a screenshot or video of the actual breakage. No "works on my machine" arguments.
PMs get functional impact, devs get stack traces and file paths — from the same analysis, without a second pass.