QA Claude Skill

32 production-grade QA workflow skills for Claude Code — from spec to release.

English · 繁體中文 · 简体中文 · 日本語

A configurable suite of 32 QA skills for Claude Code, covering the
full test lifecycle: spec → TC → automation → performance → security → review → regression → publish.
Extracted from a personal QA workspace and generalized via config.json —
drop in your team's IDs and it works in any team, any tool stack.

👀 See it in action

⚡ One-command install

💬 Natural-language skill triggers

📤 Real outputs ready to ship

✨ Highlights

• 🧪 Full lifecycle coverage — Spec parsing, TC design, automation generation, code review, regression planning, bug filing, mutation testing, and dashboard publishing
• 🔌 Tool-agnostic via 3 modes — full-mcp (Atlassian + Slack + Google) / partial-mcp (degrade gracefully when tools missing) / markdown-only (zero external dependencies)
• 🔔 Multi-channel notifications — Slack (MCP) + Telegram / Line (Bot API), can run in parallel; especially handy for Taiwan/Asia teams (Line > Slack)
• 🌐 Bilingual — Every skill ships with SKILL.md (zh-TW) + SKILL.en.md; 4 concept guides in 繁中 for unfamiliar topics
• 📦 One-config customization — 28 variables in config.json cover JIRA / Slack / Google / iOS / Android / BE pytest / AWS dashboard
• 🧩 Pluggable modules — Each skill has modules/{config-loader, jira/slack-integration, multichannel-notify, markdown-fallback}.md for clean separation
• 🚀 One-command install — ./install.sh validates config, renders 28 placeholders, backs up existing skills, installs to ~/.claude/skills/
• 🇹🇼 Designed in Taiwan, ready for global teams — Includes a11y mandatory checks (Dynamic Type / TalkBack / contrast) and cross-platform pairing baked into every skill

📦 What's in the box

32 skills across 10 categories:

Test Design (8)

Automation (3)

Bug Management (1)

Quality Quantification (2)

Reporting (2)

Performance & Security (4) — ✨ NEW in v1.5.0

CI Health (3) — ✨ NEW in v1.5.0

Quality Specialties (5) — ✨ NEW in v1.6.0

Compliance & Governance (1) — ✨ NEW in v1.7.0

Domain Specialties (3) — ✨ NEW in v1.8.0

💡 First time hearing of mutation testing / property-based testing / spec-driven dev / test tiering?
Each has a 5-minute Chinese intro at skills/<name>/concept-zh.md. See Concept Guides.

🚀 Quick start

# 1. Clone
git clone https://github.com/kao273183/qa-claude-skill.git ~/Desktop/QA_Claude_Skill
cd ~/Desktop/QA_Claude_Skill

# 2. Create your config
cp config/config.example.json config/config.json

# 3. Fill in the 4 minimum fields:
#    - jira.instance_url
#    - jira.project_key
#    - platforms.ios.default_device
#    - platforms.android.default_device

# 4. Install (renders 28 placeholders → ~/.claude/skills/)
./install.sh

# 5. In Claude Code, try a trigger phrase:
#    "Generate test plan for feature X"
#    "Write a bug report for this crash"
#    "Review these test cases"

Dry-run before installing

CLAUDE_SKILLS_DIR=/tmp/preview ./install.sh
ls /tmp/preview/   # 32 skill directories
grep -r '{{' /tmp/preview/ | grep -v '變數'   # should be empty

🎛 The 3 modes

Each skill works in all 3 modes; pick the one that matches your team's tooling:

3 ready-to-use presets ship in config/presets/ — copy one and edit:

cp config/presets/full-stack.json     config/config.json   # All MCPs
cp config/presets/jira-only.json      config/config.json   # JIRA only
cp config/presets/markdown-only.json  config/config.json   # Pure docs

⚙️ Customization

Three layers of configurability:

1. config.json — 28 variables. See docs/customization-guide.md for the full mapping.
2. config/presets/ — 3 starter scenarios (full-stack / jira-only / markdown-only)
3. Per-skill modules — Each skill has modules/markdown-fallback.md defining degraded behavior

Example configurations

• 🏢 Large team — ACME Corp — Full JIRA + Slack + Google + AWS dashboard
• 👤 Solo developer — Pure Markdown, no external deps

🧩 Architecture

Each skill follows the same pluggable structure:

skills/<skill-name>/
├── SKILL.md                          ← Main spec (zh-TW)
├── SKILL.en.md                       ← English mirror
├── concept-zh.md                     ← Beginner intro (for unfamiliar topics)
├── examples.md                       ← 3-5 real usage scenarios
├── templates.md / patterns.md        ← Templates / code patterns
└── modules/                          ← Pluggable integrations
    ├── config-loader.md              ← Load config.json values
    ├── jira-integration.md           ← (optional) JIRA MCP calls
    ├── slack-integration.md          ← (optional) Slack MCP calls
    ├── multichannel-notify.md        ← (optional) Telegram / Line via Bot API
    └── markdown-fallback.md          ← Pure Markdown degradation path

This means:

• Removing JIRA? Delete modules/jira-integration.md references — Slack still works.
• No Google? Switch to markdown-only mode — every skill stays functional.
• Adding a new tool integration? Add modules/<your-tool>.md and reference it from SKILL.md.

📖 Concept Guides

For unfamiliar testing concepts, ship-in 繁中 quick reads (5 min each):

🌊 Typical workflows

See docs/workflow-diagrams.md for ASCII diagrams of:

1. Spec → Release pipeline (BE feature) — speckit-to-tc → test-review → sheet-md-sync → tc-to-pytest → mutation-testing → property-based-test-gen
2. Pre-release prep (mobile) — test-master → test-automation → smoke-test-analyzer → regression-test → bug-report → publish-regression
3. TC version bump — test-master --quick → test-review → tc-version-diff → tc-to-pytest --incremental
4. Markdown-only flow (solo dev) — All skills produce .md under .claude/testing/
5. Tri-party review — Claude + Codex + Gemini reviewing the same TC, with weighted consensus

🧰 Compatibility

🗺 Roadmap

See ROADMAP.md for the full picture — what's shipped, what's planned, and what we're explicitly not doing.

Highlights of what's next:

• 9 more skills (test-impact-analyzer, oauth-flow-test, payment-test, graphql-test, llm-quality-eval, ...)
• Japanese translation
• Video walkthrough + demo GIF

Want to influence priority? Open a GitHub Discussion.

🔒 Security

Found a vulnerability? See SECURITY.md — preferred channel is GitHub Security Advisory.

🤝 Contributing

PRs welcome! See CONTRIBUTING.md for:

• How to add a new skill
• How to contribute translations
• How to modify existing skills
• PR checklist

📝 License & Trademark

Dual-licensed software — see LICENSE for full details:

For commercial licensing, open a GitHub Issue with commercial-license label.

Trademark: "QA Claude Skill" is a trademark of Jack Kao — see TRADEMARK.md for usage rules. The MIT/Commercial license grants source-code rights, not trademark rights. You may not name a fork "QA Claude Skill X" without permission.

This is a community / personal project for Claude Code users — NOT an official Anthropic product.

☕ Support

If this project saves your QA team time, consider buying me a coffee — it keeps the project iterating:

Or just give the repo a ⭐ — it costs nothing and helps others find it.

Made with ❤️ for QA teams who want to focus on quality, not paperwork.