Skip to main content
subagent CX Codex

qa-manual-tester — Claude Code subagent: manual browser QA

qa-manual-tester takes over a real browser and pokes at your app the way a tester would: clicks, typing, screenshots, console reads. Handy when a feature needs a hands-on pass.

no license ★ 406 updated 2026-06-25

Install into Claude Code · Manually: put the files in place All methods and platforms ↓
.claude/agents/qa-manual-tester.md
Open on GitHub

What qa-manual-tester does

A Claude subagent that drives a real browser through your app. It navigates, clicks, fills fields, picks options from dropdowns, captures screenshots and reads the console, all via the Playwright MCP tools (mcp__playwright__browser_navigate, browser_click, browser_type, browser_screenshot and the rest). It does not write test suites. The job is a manual pass, carried out by an agent.

One markdown file, model: sonnet in the frontmatter, and a prompt that fixes the order of work. Flows and their expected outcomes come first, then smoke checks and happy paths, then boundary values and error states. Anything broken gets reproduced, captured in a screenshot and rated Critical, High, Medium or Low. Reports carry the steps taken, what was expected, what happened.

Without a Playwright MCP server connected there is nothing to test: the mcp__playwright__* tools simply aren't in the session, and you get a code review where you wanted a UI check. Set that up before the first run.

What it helps with

  • check a sign-in form in the browser right after a change
  • run a regression pass over checkout following a refactor
  • reproduce a submit-button bug at a phone viewport
  • get a severity-rated defect report with screenshots
  • see what the console throws while a form submits

Example prompts for the agent

  • Run qa-manual-tester and check whether the new sign-in form works.
  • Have qa-manual-tester walk the whole checkout and note where it breaks.
  • Send qa-manual-tester to the settings page at phone width and bring back screenshots of the errors.

How to install qa-manual-tester

Manually: put the files in place

.claude/agents/qa-manual-tester.md

For all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention

Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean

Questions about qa-manual-tester

How do I install qa-manual-tester in Claude Code?

Drop the agent file at .claude/agents/qa-manual-tester.md inside your project, then call it by name.

What does qa-manual-tester do?

It drives a browser through the Playwright MCP tools: navigation, clicks, typing, screenshots, console output. You end up with a report on the defects it found and how serious each one is.

What does it need to run?

A connected Playwright MCP server, since without those tools it cannot open a page. The agent itself is set to model: sonnet.

Similar subagents

All subagents for Testing and QA →

From the same set26

The whole set →
abstraction-power The abstraction-power skill guides Claude Code from concrete code examples to reusable models by spotting their shared structure. Use it to examine duplicated logic or shape a system design. CX Codex +1 ★ 406repo stars get-to-know get-to-know is a Claude Code command that initializes project context by gathering stack, structure, and conventions, creating a project map, and configuring the environment. CX Codex +1 ★ 406repo stars adversarial-review adversarial-review subjects fresh builds to hostile inspection using isolated reviewer contexts and mandates empirical reproduction on the running system. CX Codex +1 ★ 406repo stars human-writing The human-writing skill guides an agent to shape prose around its language, voice, and format. Use it for articles, posts, headlines, and other writing tasks. CX Codex +1 ★ 406repo stars comparator comparator runs blind comparisons between two AI outputs against the original prompt. Identifies the stronger result without knowing which skill produced each file. CX Codex ★ 406repo stars code-review code-review — a subagent for Claude Code that reviews a pull request through gh and posts a comment listing confirmed bugs. Skips closed, draft, or already-reviewed PRs without prompting. CX Codex ★ 406repo stars change-core-self change-core-self asks you about the project, works out a fitting identity for the agent from scratch, and stores it as a rule file loaded on every session. CX Codex +1 ★ 406repo stars grader grader assesses agent execution transcripts and generated files against expectations, flags superficial passes, and critiques test assertions. CX Codex ★ 406repo stars

Project sponsors

Companies that keep this analytics open