Qa Skills

QA Skills for Coding Agents

A practical QA skill pack for Claude Code, Codex, Gemini CLI, OpenCode, and similar coding agents.

This repository ships 9 skills:

• 4 generators for scaffolding test suites, scenarios, and plans
• 3 analyzers for detecting issues, gaps, and risks
• 1 input tester for boundary and edge case generation
• 1 security tester for OWASP-based security test cases

Design Principles

• Native frameworks only — Tests are written using the project's own test framework (pytest, vitest, JUnit, go test). No intermediate tools like Postman, Karate, or SoapUI.
• Playwright-exclusive for E2E — All browser testing uses the Playwright ecosystem (Playwright Test, CLI codegen, MCP, Trace Viewer). No Selenium or Cypress.
• Code-first — Everything is version-controlled, CI-friendly test code. No GUI-dependent workflows.

Quick Install

Claude Code Plugin Marketplace

/plugin marketplace add umitozdemirf/qa-skills
/plugin install qa-skills@umitozdemirf

Team Rollout

Add this to project-level .claude/settings.json:

{
  "extraKnownMarketplaces": {
    "qa-skills": {
      "source": {
        "source": "github",
        "repo": "umitozdemirf/qa-skills"
      }
    }
  }
}

One-Command Installer

For local project setup across supported agents:

./scripts/install_integrations.sh all --project "$(pwd)"

Cross-Agent Install

This repo now supports three integration modes:

• Native Claude Code plugin via marketplace install
• AGENTS.md-based integration for Codex/OpenCode-style agents
• Command-pack / prompt-pack export for Claude Code, Gemini CLI, and OpenCode

Claude Code

Primary path: install the plugin from the marketplace.

Alternative path: use the installer to place slash commands into a project or user command directory.

./scripts/install_integrations.sh claude --project "$(pwd)"

This gives you commands like /qa:api-test-generator and /qa:risk-analyzer.

Codex

Codex works best here through the repository-level AGENTS.md. Keep this repo checked out inside the project, or vendor the AGENTS.md plus qa-skills-plugin/skills/ tree into a location your project instructions can reference.

Recommended project layout:

your-project/
├── AGENTS.md
└── tools/
    └── qa-skills/
        ├── AGENTS.md
        └── qa-skills-plugin/skills/

Minimal install flow:

./scripts/install_integrations.sh codex --project "$(pwd)"

This installs the QA skill pack under tools/qa-skills/ and creates a minimal root AGENTS.md if one does not already exist.

Gemini CLI

Install Gemini command wrappers with:

./scripts/install_integrations.sh gemini --project "$(pwd)"

This gives you commands like /qa:security-test-generator.

OpenCode

Install OpenCode command wrappers with:

./scripts/install_integrations.sh opencode --project "$(pwd)"

If your OpenCode setup uses .opencode/command/ instead of .opencode/commands/, copy the same qa/ folder there instead.

Install Everything

./scripts/install_integrations.sh all --project "$(pwd)"

Home Directory Install

./scripts/install_integrations.sh claude --home
./scripts/install_integrations.sh gemini --home
./scripts/install_integrations.sh opencode --home
./scripts/install_integrations.sh codex --home

Regenerate Exports

python3 scripts/export_agent_commands.py
./scripts/package_integrations.sh

How People Use This Repo

Most workflows are generator + analyzer loops.

flowchart LR
  A["Analyze risk / coverage"] --> B["Generate test plan"]
  B --> C["Generate test code"]
  C --> D["Detect test smells"]
  D --> E{"Quality OK?"}
  E -- "No" --> F["Fix + re-analyze"]
  E -- "Yes" --> G["Ship to PR/CI"]

Typical prompts:

Use api-test-generator to create pytest tests for all endpoints in src/api/
Generate service test scenarios from our swagger.json for the order flow.
Analyze risk for this PR and suggest what to test first.
Use security-test-generator to create OWASP test cases for the auth module.
Run test-smell-detector on tests/ and fix any critical issues.
Generate a test plan for the checkout feature with risk-based prioritization.

Skill Catalog (9)

Test Generation (4)

Skill	Primary use
api-test-generator	Generate API test suites using native test frameworks (pytest+httpx, vitest+supertest, JUnit+REST-assured, go test)
service-test-generator	Generate multi-step scenario tests from OpenAPI/Swagger specs — CRUD lifecycles, auth flows, cross-resource chains
e2e-test-generator	Generate Playwright E2E tests with POM, codegen, API testing, and trace support
test-plan-generator	Generate risk-prioritized test plans for features, PRs, or releases

Analysis and Detection (3)

Skill	Primary use
test-smell-detector	Detect test anti-patterns, flaky indicators, and maintainability issues
test-coverage-analyzer	Identify untested code paths, missing edge cases, and coverage gaps
risk-analyzer	Score change risk, map blast radius, identify fragile areas, prioritize testing

Specialized Testing (2)

Skill	Primary use
input-validation-tester	Generate boundary value, equivalence partitioning, and edge case test data
security-test-generator	Generate OWASP Top 10 security test cases for web apps and APIs

Supported Frameworks

API Testing

Language	Framework	HTTP Client
Python	pytest	httpx, requests
JavaScript/TypeScript	vitest, jest	supertest, fetch
Java	JUnit 5	REST-assured, WebTestClient
Go	go test	net/http/httptest

E2E Testing

Capability	Tool
Browser testing	Playwright Test
Test recording	Playwright CLI codegen
API within E2E	Playwright request context
Component testing	Playwright experimental CT
Debugging	Playwright Trace Viewer, UI Mode
AI-assisted	Playwright MCP

Repo Layout

qa-skills/
├── AGENTS.md
├── .github/workflows/validate.yml
├── README.md
├── LICENSE
├── .gitignore
├── scripts/
│   ├── export_agent_commands.py
│   ├── install_integrations.sh
│   ├── package_integrations.sh
│   └── validate_skills.py
├── integrations/
│   ├── claude/
│   ├── gemini/
│   ├── opencode/
│   └── README.md
├── .claude-plugin/
│   └── marketplace.json
└── qa-skills-plugin/
    ├── .claude-plugin/plugin.json
    └── skills/
        └── <skill-name>/
            ├── SKILL.md
            ├── references/
            ├── scripts/        # optional
            ├── examples/       # optional
            ├── test/           # optional
            └── assets/         # optional

Requirements

Baseline

• bash
• python3 (3.9+ recommended)
• node (18+ for Playwright)

By Domain

Domain	Tools
API testing (Python)	pytest, httpx, faker
API testing (JS/TS)	vitest or jest, supertest
E2E testing	@playwright/test, Playwright browsers
Security testing	Application under test, curl
Coverage	pytest-cov, c8/istanbul, jacoco

Quick Install (macOS)

# Python API testing
pip install pytest httpx pytest-cov faker

# Node.js API testing
npm install -D vitest supertest

# Playwright E2E
npm init playwright@latest
npx playwright install

# All-in-one
brew install python node

Contributing

Contributions are welcome for:

• New skills in adjacent QA domains (accessibility, contract testing, performance)
• Better reference coverage and examples
• Scripts for automated detection and analysis
• CI integration examples

Examples

Example outputs are now included for high-value skills:

• api-test-generator example
• service-test-generator example
• e2e-test-generator example
• input-validation-tester example
• test-plan-generator example
• risk-analyzer example
• security-test-generator example
• test-coverage-analyzer example
• test-smell-detector example

Validation

Validate the skill pack and regenerate cross-agent command wrappers with:

python3 scripts/validate_skills.py
python3 scripts/export_agent_commands.py
./scripts/package_integrations.sh

Agent Compatibility Notes

• Claude Code: native plugin install remains the cleanest path; exported slash commands are useful when you want local project commands instead.
• Codex: this repo ships a root AGENTS.md so the skill catalog can be consumed as project instructions.
• Gemini CLI: wrappers are exported as TOML command files under integrations/gemini/.
• OpenCode: wrappers are exported as markdown command files under integrations/opencode/.

Source Links

• Claude Code slash commands
• OpenAI Codex repository
• Gemini CLI custom commands
• OpenCode commands

License

Apache-2.0