umitozdemirf

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)