Skip to content

impravin22/my-claude-maestro

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

72 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Maestro — Master Orchestrator for Claude Code

A Claude Code plugin that orchestrates your entire development workflow. Maestro activates at the start of every task to classify work, fetch live library documentation, enforce engineering standards, and guide you through a disciplined build-verify-ship cycle.

What It Does

  1. Classifies your task (feature, bug fix, refactor, config, UI-only)
  2. Fetches live docs via Context7 for every library involved — no stale training data
  3. Orchestrates superpowers skills in the correct order (brainstorm → plan → TDD → implement → verify → PR)
  4. Enforces UI/UX design system — WCAG 2.1 AA accessibility, Tailwind token usage, shadcn/ui patterns, responsive design, loading/error/empty states. Step 5 runs an anti-template design-mockup gate that produces an approved visual artefact (HTML prototype, sketch, or Storybook story) before any production frontend code is written. The gate enforces an explicit anti-template ban (no centred max-w-md card with icon→headline→CTA, no "clean minimal", no unmodified Tailwind defaults), requires the design to demonstrate at least four of ten quality markers (hierarchy, rhythm, depth, typography, semantic colour, drawn states, grid-breaking, atmosphere, motion, dataviz), and self-audits before the user is asked to approve — eliminating both the post-implementation rework loop and the template-by-default failure mode
  5. Enforces layered security — OWASP checklists at planning time + real-time pre-edit scanning via Security Guidance
  6. Visual verificationPlaywright MCP verifies frontend changes render correctly, pass accessibility checks, and behave across breakpoints
  7. Deep PR reviewPR Review Toolkit dispatches specialist agents (code review, silent failure detection, test coverage, type design, code simplification, comment accuracy) before the polling loop
  8. Cross-session memoryclaude-mem surfaces prior observations (decisions, rejected approaches, failed experiments) during CLASSIFY, BRAINSTORM, and PLAN via the search, timeline, and get_observations MCP tools — no more re-deriving context that already exists
  9. Composes with an extended plugin ecosystemUI UX Pro Max for design styles + palettes, n8n-MCP for 400+ n8n integrations, VoiceMode MCP for voice conversations, Everything Claude Code for 150+ skills across 12 language ecosystems, and LightRAG as an optional graph+vector RAG supplement
  10. Enforces quality gates — tests mandatory, lint clean, format clean, TypeScript clean, solution justification, British English
  11. Tracks upstream dependencies — daily GitHub Actions workflow detects changes in all 12 tracked upstream repos, files issues, and auto-creates PRs for review

Prerequisites

Dependency Required Install
superpowers plugin Yes Comes with Claude Code
Context7 MCP Yes npx ctx7 setup --claude
Vercel plugin Recommended Provides shadcn + react-best-practices skills
Security Guidance Recommended /plugin install security-guidance@anthropic — real-time pre-edit security scanning
PR Review Toolkit Recommended Ships with Claude Code — 6 specialist review agents
Playwright MCP Recommended npx @anthropic-ai/claude-code mcp add playwright -- npx @anthropic-ai/mcp-playwright
claude-mem Recommended npx claude-mem install — persistent memory across sessions via 5 lifecycle hooks + 3 MCP tools (search, timeline, get_observations)
frontend-design skill Recommended Ships with Everything Claude Code (or any equivalent) — used by Step 5a/5b to generate the design direction and mockup artefact before any production frontend code is written
UI UX Pro Max Recommended npm i -g uipro-cli && uipro init --ai claude — 50+ styles, 161 colour palettes, 99 UX guidelines (auto-activates on UI/UX prompts; Step 5e checklist remains canonical)
n8n-MCP Recommended claude mcp add n8n-mcp -e MCP_MODE=stdio -e LOG_LEVEL=error -e DISABLE_CONSOLE_OUTPUT=true -- npx -y n8n-mcp — 400+ n8n workflow integrations
VoiceMode MCP Recommended claude mcp add --scope user voicemode -- uvx --refresh voice-mode — local Whisper + Kokoro voice conversations (requires mic/speakers)
Everything Claude Code Recommended git clone https://github.com/affaan-m/everything-claude-code.git && cd everything-claude-code && ./install.sh --target claude --profile full — 150+ skills, 47 agents, 79 commands, 16 rules across 12 language ecosystems
LightRAG Recommended uv tool install "lightrag-hku[api]" — graph+vector RAG Python library; optional supplement to Context7 for large codebases (external service; custom MCP bridge required to surface inside Claude Code)
Andrej Karpathy Skills Recommended claude plugin marketplace add forrestchang/andrej-karpathy-skills && claude plugin install andrej-karpathy-skills@karpathy-skills — Karpathy's 4 LLM-coding principles (think before coding, simplicity first, surgical changes, goal-driven execution) as an enforced voice that composes with maestro's own engineering-mindset discipline
Caveman Recommended claude plugin marketplace add JuliusBrussee/caveman && claude plugin install caveman@caveman — ultra-compressed communication mode that cuts ~75% token usage while preserving full technical accuracy

Installation

For Individual Use

1. Install Maestro itself:

/plugin marketplace add impravin22/my-claude-maestro
/plugin install maestro@impravin22

2. Install the companion ecosystem (recommended):

# Clone the repo for the bundled installer
git clone https://github.com/impravin22/my-claude-maestro.git
cd my-claude-maestro
./install.sh

The installer handles: superpowers, Context7 MCP, Vercel plugin, Security Guidance, PR Review Toolkit, Playwright MCP, claude-mem, UI UX Pro Max, Andrej Karpathy Skills, Caveman, and Everything Claude Code.

Heavy/specialised dependencies (VoiceMode, n8n-MCP, LightRAG) are excluded by default — install manually from the Prerequisites table below if you need them.

Installer flags:

./install.sh --minimal          # required components only (superpowers + Context7)
./install.sh --dry-run          # preview commands without executing
./install.sh --skip-vercel      # opt out of individual components
./install.sh --help

Restart Claude Code after installation.

For Team-Wide Enforcement

Add to your team's .claude/settings.json:

{
  "extraKnownMarketplaces": {
    "impravin22": {
      "source": {
        "source": "github",
        "repo": "impravin22/my-claude-maestro"
      }
    }
  }
}

Then each team member runs:

/plugin install maestro@impravin22

The Unified Flow

Every task follows one flow. Steps are skipped when not applicable:

 1. CLASSIFY     → Determine task type and scope
 2. CONTEXT7     → Detect libraries → fetch current docs
 3. BRAINSTORM   → superpowers:brainstorming (or systematic-debugging for bugs)
 4. PLAN         → superpowers:writing-plans
 5. UI/UX GATE   → Generate approved design mockup → full design system checklist (frontend only)
 6. SECURITY     → OWASP + LLM security checklist + real-time edit scanning
 7. IMPLEMENT    → superpowers:test-driven-development
 8. VERIFY       → superpowers:verification + quality gates + Playwright visual checks
 9. FINISH       → superpowers:finishing-a-development-branch → PR
10. REVIEW       → PR Review Toolkit specialist agents → polling loop

Skip Logic

Condition Steps Skipped
Trivial config/docs change 3–6
No frontend touched 5, visual verification in 8
Bug fix 3 → systematic-debugging
No libraries detected 2
No dev server running Visual verification in 8
No new types introduced type-design-analyzer in 10
No comments added/modified comment-analyzer in 10

Checklists

The plugin includes three comprehensive reference checklists:

  • UI/UX Design System — visual design, accessibility, component patterns, performance, user workflow
  • Security (OWASP) — injection, auth, access control, input/output protection, LLM security, dependencies
  • Quality Gates — testing, linting, code quality, visual verification (Playwright), PR specialist review, solution justification, style, git workflow

Customisation

The checklists in skills/maestro/references/ are plain Markdown. Fork the repo and edit them to match your team's standards:

  • Add or remove checklist items
  • Change tool-specific commands (e.g., swap Vitest for Jest)
  • Adjust accessibility level (WCAG 2.1 AA → AAA)
  • Add project-specific security rules

External Resources

Curated references — not integrations, but useful while working with Claude Code. Link-only due to licence restrictions (cannot be vendored into an MIT-licensed repo):

Plugin Structure

my-claude-maestro/
├── .claude-plugin/
│   ├── plugin.json
│   └── marketplace.json
├── skills/
│   └── maestro/
│       ├── SKILL.md
│       └── references/
│           ├── uiux-checklist.md
│           ├── security-checklist.md
│           └── quality-gates.md
├── hooks/
│   ├── hooks.json
│   └── check-update.sh
├── docs/
│   ├── 2026-04-03-maestro-design.md
│   ├── 2026-04-07-plugin-integration-design.md
│   ├── 2026-04-13-claude-mem-integration-design.md
│   └── 2026-04-13-multi-plugin-integration-design.md
├── install.sh          # companion ecosystem installer
├── README.md
└── LICENSE

Licence

MIT

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages