A comprehensive, production-ready Claude Code configuration for Laravel + Inertia.js + Vue 3 projects. Includes 17 specialized agents, 10 rule files, 23 skills, and a structured workflow pipeline that turns Claude Code into a full AI development team.
Stack: PHP 8.4 · Laravel 12 · Vue 3 · Inertia.js v2 · PostgreSQL 17 · Redis · Docker · Pest 4
Specialized AI agents that handle different aspects of development. Each agent has an explicit tools: list in its frontmatter — only the tools it actually needs (principle of least privilege).
| Agent | Purpose | Model | Write? | Key extras |
|---|---|---|---|---|
ba |
Business analysis, requirements, user stories | opus | — | Web, Context7, Agent |
dba |
Database design, migrations, query optimization | sonnet | + | — |
ddd-architect |
Domain modeling, business logic placement | opus | — | Context7, Agent |
debugger |
Bug investigation, root-cause analysis | opus | + | — |
developer |
Full-stack Laravel + Inertia.js features | sonnet | + | Context7, Figma, IDE, Agent |
devil |
Devil's advocate in planning phase, challenges requirements and architecture | opus | — | SendMessage only |
devops |
Docker, CI/CD, deployment, GitHub Actions, infrastructure | haiku | + | GitHub MCP |
docs-writer |
Technical documentation, README, API docs | haiku | + | GitHub MCP |
filament |
Filament v4 admin panel resources | sonnet | + | Context7 |
frontend |
Vue 3 components, Pinia, Tailwind, a11y | sonnet | + | Context7, Figma, Stitch, IDE |
integration-architect |
OAuth, webhooks, third-party services | sonnet | + | Web, Context7 |
laravel-refactoring-expert |
Refactoring, N+1 fixes, code quality | sonnet | + | — |
qa |
E2E testing, Playwright, visual regression | sonnet | + | All 21 Playwright tools |
queue-specialist |
Redis queues, jobs, async processing | sonnet | + | — |
reviewer |
Code review, architecture audit | sonnet | — | GitHub MCP (review) |
security-scanner |
OWASP, auth/authz, credential leaks | opus | — | Web (CVE lookup) |
tester |
Unit/feature tests, Pest, mutation testing | sonnet | + | — |
Rule files loaded by agents and orchestrator on demand:
| Rule | Purpose |
|---|---|
architecture.md |
Actions pattern, Inertia.js, domain organization |
code-style.md |
PHP 8.4 strict types, Eloquent conventions, Pint/PHPStan/Rector |
docker-commands.md |
Docker-prefixed commands reference |
forms-authorization.md |
Form Request + Policy + authorization patterns |
git-operations.md |
Commit/push safety, PR description format |
inertia-vue.md |
Inertia v2 + Vue 3 Composition API conventions |
mcp-stack.md |
MCP tool usage guide (Laravel Boost, Context7, GitHub, Figma) |
migrations-queue.md |
Migration conventions, AsJob queue pattern |
testing.md |
Pest 4, mutation testing, model testing policy |
workflow.md |
Agent pipeline orchestration + agent routing table |
Three files (workflow.md, code-style.md, git-operations.md) are auto-imported in CLAUDE.md via @-imports — always in context. The rest are loaded by agents on demand via reference links.
Reusable knowledge modules organized by category:
Laravel & PHP: laravel-architecture, php-pro, laravel-actions-patterns (custom)
Testing: pest-testing, test-master, playwright-expert, playwright-skill
Database: database-optimizer, postgresql, postgres-best-practices
Architecture: architecture-designer, ddd-strategic-design, code-reviewer, architect-review
DevOps: devops, docker-expert, github-actions
Debugging & Security: debugging-wizard, security-reviewer
Frontend: vue-expert
Infrastructure: octane-frankenphp-gotchas (custom)
Planning: plan-writing, brainstorming
The configuration defines a mandatory agent pipeline for feature development:
Standard Feature: Planning Team → Developer → Quality Gate Team → DocsWriter
Bug Fix: Debugger → Developer → Verify Team
CI/CD: DevOps → Reviewer + Security Scanner
Planning Team (plan-{slug}) runs ba, ddd-architect, and devil in parallel. devil is a read-only devil's advocate that challenges requirements and architecture decisions via SendMessage before any code is written. For simple features with no arch decisions, ba runs sequentially alone.
Quality Gate Team (qg-{slug}) runs tester, reviewer, security-scanner, and qa in parallel via TeamCreate. If any agent reports a Critical or Important issue, findings route back to Developer and the team reruns.
- Claude Code CLI installed
- Node.js 18+ (for skills installation via
npx) - A Laravel project (ideally with Docker)
Clone this repository and copy the configuration files into your Laravel project:
# Clone the config repo
git clone https://github.com/AlexGritsworker/claude-laravel.git /tmp/claude-laravel-config
# Copy into your project
cp -r /tmp/claude-laravel-config/.claude /path/to/your-laravel-project/
cp /tmp/claude-laravel-config/CLAUDE.md /path/to/your-laravel-project/Or add as a git subtree/submodule if you prefer to track updates.
Agent Teams:
settings.jsonalready includesCLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1which enables parallel Quality Gate and Verify Team execution. This is an experimental feature — requires Claude Code with agent teams support.
Superpowers provides structured development workflows: brainstorming, planning, TDD, debugging, code review, and more.
Run inside Claude Code:
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace
Skills provided: brainstorming, writing-plans, executing-plans, test-driven-development, systematic-debugging, requesting-code-review, verification-before-completion, using-git-worktrees, subagent-driven-development, finishing-a-development-branch
The configuration already includes 27 project skills in .claude/skills/. To install additional community skills from skills.sh:
npx skills add <owner/repo>Browse the skills.sh leaderboard to discover popular skills. Examples:
# Install a specific skill set
npx skills add vercel-labs/skills
# Install with options
npx skills add <owner/repo> -a claude-codeThe main instructions file. Adapt this to your project:
- Update system requirements if your stack differs
- Adjust Docker commands to match your
compose.yml - Add project-specific rules in the "Core Principles" section
- Modify the agent list if you add/remove agents
Create .claude/agents/my-agent.md:
---
name: my-agent
description: "What this agent does. Trigger words — EN: keyword1, keyword2. Trigger words — UA: слово1, слово2."
model: sonnet
color: blue
tools:
- Read
- Glob
- Grep
- Edit
- Write
- Bash
- SendMessage
---
# Agent Title
Instructions for the agent...Always define tools: explicitly — agents inherit ALL tools from the conversation if the field is omitted. Grant only what the agent needs:
- Read-only agents (analysts, auditors): omit
Edit,Write,Bash - Agents that don't call external services: omit MCP tools
qais the only agent that should get Playwright MCP toolsreviewer,devops,docs-writerare the only agents that need GitHub MCP
Create .claude/rules/my-rule.md with markdown content. Rules are auto-loaded for every conversation.
Create .claude/skills/my-skill/SKILL.md:
---
name: my-skill
description: What this skill does and when to use it
---
Instructions Claude follows when this skill is active...See the Claude Code Skills documentation for advanced features: supporting files, subagent execution, dynamic context injection.
All agents include trigger words in both English and Ukrainian. To add another language, extend the description field in the YAML frontmatter:
Trigger words — EN: keyword1, keyword2.
Trigger words — UA: слово1, слово2.
Trigger words — DE: schlüsselwort1, schlüsselwort2.
This configuration follows the Laravel Actions pattern (lorisleiva/laravel-actions):
| Layer | Pattern |
|---|---|
| HTTP entry | Page Actions (AsController) |
| Form handling | Store/Update Actions (AsController) |
| Business logic | Business Actions (AsObject) |
| Authorization | Policies |
| Validation | Form Requests |
| Side effects | Observers |
| Value objects | Enums |
| Async work | Jobs (ShouldQueue) |
| Cross-cutting | Events / Listeners |
No traditional Controllers, no Repository pattern, no app/Domain/ directory.
MIT


