deep-review

Proportione Plugins — Claude Code Playbook

One-liner install:

claude plugin marketplace add Proportione/proportione-plugins

---

How a 1-3 person IT consultancy uses Claude Code to manage 13 WordPress sites, 4 GCP microservices, a scientific monorepo, and IaC infrastructure — with quality.

What's here

This repo is both a Claude Code plugin marketplace (installable skills) and a playbook of everything we've learned using Claude Code in production since early 2026.

Plugins (9 QA skills)

Install the marketplace, then pick what you need:

claude plugin marketplace add Proportione/proportione-plugins
claude plugin install deep-review@proportione-plugins

Plugin	What it does
deep-review	Deep code review for PRs, commits, or files — logic, naming, complexity, tests, security
check-security	OWASP-based security audit with Semgrep and manual analysis
check-architecture	Validate code against architectural conventions defined in CLAUDE.md
check-performance	Performance analysis: N+1 queries, bundle size, Web Vitals, Lighthouse, GCP sizing
check-wordpress	Full WordPress QA: accessibility, SEO, links, images, security, performance
check-wp-visual	Visual regression testing for WordPress sites using Playwright
check-terraform	IaC validation: terraform validate/plan + Checkov + Semgrep
triage-bug	Automated bug triage: error analysis, root cause, fix, and PR creation
update-whitepaper	Monthly QA whitepaper update with metrics collection and source research

Playbook (guides)

Hard-won lessons from running Claude Code in a real production environment:

Guide	Topic
CLAUDE.md as Contract	How we use CLAUDE.md as an architectural contract, not documentation
MCP Servers	Building MCP servers for ticketing, CRM, BI — and keeping auth working
Playwright Tricks	Automating Google services, Synology DSM, Looker Studio, and more
QA Framework	The five-layer QA model for AI-generated code
Operational Rules	Governance rules for Claude Code in automation pipelines
GCP Patterns	Cloud Run, Firestore, BigQuery patterns we use daily
Sales & Marketing	CRM, prospecting, content, social media — all from the terminal
Spec-Driven Ecosystem	GSD, Superpowers, SuperClaude, Alfred Dev — how spec tools complement QA

Our stack

• WordPress (13 sites): Elementor Pro + SiteGround
• Backend: Node.js 20 (Cloud Run), Python 3.14 (FastAPI), Terraform + Ansible
• GCP: Cloud Run, Firestore, BigQuery, Cloud Tasks, Vertex AI
• QA tools: Semgrep, Checkov, Lighthouse, Playwright, axe-core
• IA stack: Claude Code (architect) + Qwen 2.5 Coder 14B local (routine tasks)

Why publish this

We wrote a whitepaper arguing that AI-generated code needs AI-assisted QA. These are the tools we built to back that argument. If you run a small team writing code with Claude, these skills and patterns might save you time.

Community

• Discussions — questions, ideas, use cases
• Issues — bugs, feature requests
• Contributing — how to propose changes
• Security policy — how to report vulnerabilities
• Code of conduct

License

MIT

---

Built by Proportione — strategy, technology, and people.