AI-Assisted QA & Manual Test Checklists

Use agents to generate test cases and checklists while keeping real-device manual QA in your hands.

Agents are good at imagining edge cases and bad at seeing your screen. Use them to widen your thinking, then verify on a phone.

Core concepts

  1. Edge-case generation: Ask for tests across empty, loading, error, offline, long text, small screens, rotation and interrupted flows. — Cursor docs
  2. Manual QA checklist: Keep a checklist file in the repo and walk it before every release, on at least two devices. — React Native debugging
  3. Accessibility checks: Turn on VoiceOver and TalkBack, increase font size and verify contrast and tap targets. — React Native accessibility
  4. Bug reports for agents: Give agents exact steps, the expected result, the actual result, the device and the log. Vague bug reports waste time. — Claude Code docs

Concept videos

YouTube channels and searches for these topics. On a large screen they open in the reading panel next to the roadmap.

Explanation

Real device nights

Schedule an hour before every release to use the app like a stranger: fresh install, poor network, interruptions.

Close the loop

Every bug you find becomes a test, so it cannot come back.

What Is It?

AI-assisted QA uses agents to design tests and checklists, with humans running real-device checks.

Why Should I Care?

It finds cases you did not think of while keeping judgement with a person.

Real World Analogy

The agent drafts the inspection form; you walk the building with it.

Visual Mental Model

text
spec -> agent edge cases -> checklist -> real device run -> bugs -> new tests

Smallest Possible Example

text
Release QA checklist
- Fresh install, sign up, onboarding
- Airplane mode: open, create, edit
- Kill app mid-save, relaunch
- Large font size, dark mode, small phone
- Deny each permission, then grant in settings

Common Mistakes

❌ Trusting an agent's 'all tests pass' without running them.

Run the commands yourself and read the output.

❌ Fixing bugs without adding a test.

The bug will return.

Build This Mini Project

Create QA.md with 15 checks, run it on two devices, and turn the first three bugs into automated tests.

Prompts for Cursor, Claude Code & Codex

Paste these into Cursor, Claude Code, Codex, or any coding agent. Replace the bracketed parts with your app details, and keep the agent in plan mode first so you review the approach before it edits files.

Edge-case matrix

text
For each screen in SPEC.md, produce a table of edge cases: empty, loading, error, offline, long text, small screen, interrupted flow, permission denied. Mark which should become automated tests.

Triage a bug

text
Bug: [steps, expected, actual, device, log]. Reproduce with a failing test, explain the root cause in two sentences, fix it, and confirm the test passes.

Rules, databases & blogs to use

Resources

Related reads on iHateReading

Threads in the same domain as this chapter — go deeper on iHateReading without leaving the roadmap.

Making AI applications without AI slop cover

Making AI applications without AI slop

Quality and safety guardrails for AI-built software.

Cursor, Codex, Claude Code prompts collection cover

Cursor, Codex, Claude Code prompts collection

200+ copy-paste prompts for coding agents to build better products.