Claude Code Plugin  ·  Free & Open Source

QA Toolkit for Claude Code
Test the code. Trust the release.

15 QA commands for Claude Code — test cases, PR reviews, bug reports, and more. Auto-detects your stack. Zero config.

$ claude plugin add github:cyberwalk3r/qa-toolkit
0 slash commands
0 QA agents
0 config required
0 tech stacks

scroll to see all 15 commands

Bug reports in seconds. Describe what broke. Get back severity classification, steps to reproduce, expected vs actual, and environment fields — formatted for your tracker, not a generic template.

PR reviews that actually help. It reads the diff, maps which files touch which user flows, and writes a QA checklist grounded in what actually changed — not a checklist you've seen a hundred times.

Artifacts that fit your stack. It knows you're using Jest, not Pytest. It knows your CI runs on GitHub Actions. The output matches what your team already does, automatically.

Three steps. That's it.

Install once. Open any project. It scans your stack before you type your first command.

01

Install the plugin

One command. Works with any project — React, Django, .NET, Go, whatever you're building.

02

Open your project

On session start, QA Toolkit scans your stack and saves context. Knows your frameworks, test tools, and CI setup automatically.

03

Describe what you need

Type a slash command in plain English. Because it already knows your stack, the output uses the right format, the right frameworks, the right test runner.

Skills chain together. Each one feeds the next.

Start anywhere — every command works standalone. But chaining builds richer, state-aware context.

graph LR
    setup["setup"] --> prreview["pr-review"]
    prreview --> regplanner["regression-planner"]
    regplanner --> releaseready["release-readiness"]
    prreview --> bugreport["bug-report"]
    regplanner --> testcases["test-cases"]
    testcases --> e2etest["e2e-test"]
    testcases --> testdata["test-data"]
      

Run /qa-toolkit:setup first to confirm your detected stack, then follow the Suggested Next Steps at the bottom of each artifact.

15 slash commands. Every QA workflow covered.

Grouped by where they fit in your workflow. Each one produces a formatted artifact saved to qa-artifacts/.

Authoring
/qa-toolkit:test-cases

Test cases from requirements — table, Gherkin, or checklist format

/qa-toolkit:test-plan

Full test plan with scope, strategy, and coverage targets

/qa-toolkit:test-data

Synthetic test data — JSON, CSV, or SQL

/qa-toolkit:e2e-test

Playwright test scaffold with line-by-line comments

/qa-toolkit:api-test

API test suite — cURL, Postman collection, or Playwright

Review
/qa-toolkit:pr-review

Risk-flagged PR review with plain-English summary and QA checklist

/qa-toolkit:bug-report

Structured bug report from a casual description

/qa-toolkit:exploratory-testing

Exploratory test charters with session-based heuristics

Release
/qa-toolkit:regression-planner

Risk-based regression plan with time estimates

/qa-toolkit:release-readiness

Go/no-go assessment with quality gate scoring

/qa-toolkit:risk-prioritization

Ranked risk matrix across features and change areas

Analysis
/qa-toolkit:coverage-gap

Coverage gap analysis against requirements or test plan

/qa-toolkit:flaky-test-diagnosis

Root cause analysis for flaky tests with fix recommendations

/qa-toolkit:accessibility

WCAG 2.1 audit with plain-English manual test scripts

/qa-toolkit:setup

Read project docs, confirm detection, save preferences

When you need a conversation, not a document.

Three QA personas for multi-turn interactive work. They remember context across the conversation and adapt as you share more.

qa-reviewer

The Code Reviewer's QA Brain

Walk me through this PR. What should I actually be testing? Translates code changes into testing impact in plain English.

qa-explorer

The Edge Case Hunter

Help me find what could break in this feature. Generates exploratory test charters and surfaces edge cases you wouldn't think to test.

qa-lead

The Release Decision Maker

Let's decide if we're ready to ship. Makes release decisions, plans regression scope, and produces executive summaries.

Built for real QA workflows

  • State-aware skills — every command reads your detected project context
  • 15 skills covering authoring, review, release, and analysis
  • 3 agents with explicit tool boundaries, typed return contracts, and persistent memory
  • Multi-format output — artifacts in the format that actually fits the deliverable
  • Shared references — consistent behavior across all skills

One command to install. Zero to configure.

Works with any project. Configurable if you want — but you don't have to.

# From GitHub claude plugin add github:cyberwalk3r/qa-toolkit # Local development claude plugin add ./qa-toolkit

On first session, auto-detection runs and saves your project context to qa-artifacts/.qa-context.json. Every subsequent command reads from it.