magic-powers.pmai.space — A Claude Code plugin with cost-optimized model routing, 43 built-in skills + 67 optional skills + 42 cloud division skills (216 total), 11 specialized agents, 12 slash commands, and optional hooks/MCP integration. Built for all professional roles — dev, product, data, marketing, sales, design, and more. Also works with Cursor, Copilot, Aider, Windsurf, Gemini CLI, Codex, Kiro, and OpenCode.
Claude Code defaults to using the most expensive model for everything. Magic Powers routes tasks to the right model — ~75% cost reduction with no quality loss on routine tasks.
For actual API-level savings, combine with Claude Code Router or OpenRouter for 80-95% cost reduction.
# Claude Code — marketplace (recommended)
/plugin marketplace add kienbui1995/magic-powers
/plugin install magic-powers@magic-powers
# skills.sh (works with 43+ agents)
npx skills add kienbui1995/magic-powers
# All other tools — npx (auto-detects your tool)
npx magic-powers@latest
# Or with curl
curl -fsSL https://raw.githubusercontent.com/kienbui1995/magic-powers/main/scripts/get.sh | bashThe installer auto-detects Cursor, Windsurf, Kiro, Codex, OpenCode, Gemini CLI, and Claude Code from your environment.
git clone https://github.com/kienbui1995/magic-powers.git
cd magic-powers && bash scripts/install.sh
# Select option 7 — Codexgit clone https://github.com/kienbui1995/magic-powers.git
cd magic-powers && bash scripts/install.sh
# Select option 8 — KiroMagic Powers uses a two-tier architecture:
Everything works out of the box — no setup required:
- 12 slash commands —
/brainstorming,/plan,/review,/debug,/deploy, etc. - **12 agents** — architect, debugger, reviewer, ui-designer, etc.
- 43 built-in skills — brainstorming, TDD, debugging, security review, etc.
- SessionStart hook — auto-loads skill routing on every session
Run /setup to personalize for your project. Detects your stack, asks your role & priority, then offers:
| Feature | Type | Description |
|---|---|---|
| 🛡️ Safety guard | Hook | Block writes to .env, secrets, node_modules |
| 🔍 Auto-lint | Hook | Run linter after every file edit |
| 🧪 Auto-test | Hook | Run related tests after code changes |
| 🎨 Stitch Design | MCP | Generate UI designs (needs STITCH_API_KEY) |
| 📚 Context7 | MCP | Fetch latest library docs automatically |
| 📋 Project conventions | Skill | Coding rules based on detected stack |
| 🏗️ Stack-aware agents | Agent | Architect & debugger with stack context |
| 📦 Optional skill packs | Skills | 67 role-specific skills across 11 categories |
Optional features install to .claude/ in your project — they don't affect the global plugin.
Browse and install with /install-skills. Quick role-based install during /setup.
| Category | Skills | Who it's for |
|---|---|---|
| 🎯 Product | 6 | Product managers, founders |
| 📊 Data/ML | 9 | Data scientists, ML engineers |
| ⚙️ Platform/SRE | 6 | SRE, platform, DevOps |
| 🎨 Design/UX | 4 | Product designers, UX researchers |
| 👥 Team Processes | 9 | Team leads, eng managers |
| 📣 Marketing | 8 | Marketers, growth engineers |
| 💼 Sales | 5 | Sales, BD, account executives |
| 🧪 Testing | 5 | QA engineers, SDETs |
| 🎮 Game Dev | 5 | Game designers, developers |
| 🌐 Spatial Computing | 3 | XR/AR/VR, visionOS developers |
| 🔬 Specialist | 7 | Legal, finance, devrel, solutions arch |
See docs/OPTIONAL_SKILLS.md for the full catalog.
Professional certification-aligned agents for GCP, AWS, and Azure. /install-skills → Category 12.
| Division | Agents | Certifications |
|---|---|---|
| GCP | 7 | Professional Data Engineer, Cloud Developer, Network, ML, DevOps, Security, Architect |
| AWS | 7 | Data Engineer (DEA-C01), Developer, Networking, ML Engineer (MLA-C01), DevOps, Security, Solutions Architect |
| Azure | 7 | Fabric Data Engineer (DP-700), AI Developer (AI-200), Network, AI Engineer, DevOps, Security (SC-500), Architect |
/install-skills → 12 → GCP / AWS / Azure
@gcp-data-engineer Design a BigQuery streaming pipeline
@aws-solutions-architect Review our multi-account architecture
@azure-ai-engineer Build a RAG system with Azure OpenAI
Product analytics agents powered by the Amplitude MCP server. /install-skills → Category 13.
Requires: Amplitude MCP configured.
| Agent | Role |
|---|---|
@amplitude-analyst |
Charts, dashboards, daily/weekly briefings |
@amplitude-experimenter |
A/B tests, opportunity discovery, user journeys |
@amplitude-engineer |
Analytics instrumentation, event taxonomy |
@amplitude-ux-researcher |
Session replay, error diagnosis, UX audit |
@amplitude-ai-monitor |
AI/LLM quality, topic analysis, session investigation |
Skills ported from amplitude/mcp-marketplace.
| Command | Skill | Purpose |
|---|---|---|
/setup |
— | Personalize plugin, install optional features (12 roles) |
/install-skills |
— | Browse and install optional skills by category |
/brainstorming |
brainstorming | Explore ideas → design → spec → approval |
/plan |
writing-plans | Break feature into implementation steps |
/review |
requesting-code-review | Quick code review for bugs & issues |
/debug |
systematic-debugging | Reproduce → isolate → fix → verify |
/security-scan |
security-review | Scan for vulnerabilities & auth issues |
/db-review |
database-optimization | Schema, queries, indexes, migrations |
/deploy |
ci-cd-pipeline | Deployment readiness checklist |
/pr |
pr-workflow | Prepare changes for pull request |
/refactor |
refactoring | Improve structure, reduce complexity |
/tdd |
test-driven-development | Red → green → refactor cycle |
| Agent | Emoji | Model | Purpose |
|---|---|---|---|
architect |
🏗️ | Opus | Brainstorming, system design, planning |
debugger |
🐛 | Sonnet | Systematic debugging with full tool access |
database-optimizer |
🗄️ | Sonnet | Schema review, query optimization, migrations |
sre |
🔧 | Sonnet | Infrastructure, deployment, reliability |
git-workflow |
🌿 | Sonnet | Branch strategy, commits, release workflows |
| Agent | Emoji | Model | Purpose |
|---|---|---|---|
ui-designer |
🎨 | Sonnet | Frontend design with Stitch SDK |
| Agent | Emoji | Model | Purpose |
|---|---|---|---|
reviewer |
👀 | Haiku | Fast code review (read-only) |
security-reviewer |
🛡️ | Haiku | Security audit, vulnerability scanning |
technical-writer |
📝 | Haiku | Documentation, READMEs, ADRs, changelogs |
| Agent | Emoji | Model | Purpose |
|---|---|---|---|
product-strategist |
📣 | Sonnet | Feature prioritization, PRDs, launch planning |
copywriter |
✍️ | Haiku | Landing pages, marketing copy, announcements |
| Model | Agents | Cost |
|---|---|---|
| Opus | 1 (architect) | $$$$$ |
| Sonnet | 6 (debugger, db-optimizer, sre, git-workflow, ui-designer, product-strategist) | $$ |
| Haiku | 4 (reviewer, security-reviewer, technical-writer, copywriter) | $ |
Core Workflow: using-magic-powers · brainstorming · writing-plans · executing-plans
Planning: spec-driven-development · mvp-rapid-development
Development: test-driven-development · systematic-debugging · verification-before-completion · refactoring · api-design · performance-optimization · environment-setup
DevOps: ci-cd-pipeline · docker-containerization · incident-response · dependency-management
Collaboration: requesting-code-review · receiving-code-review · subagent-driven-development · dispatching-parallel-agents · pr-workflow
Git: using-git-worktrees · finishing-a-development-branch
Security & Auth: security-review · authentication-patterns · accessibility-compliance
Infrastructure: database-optimization · infrastructure-review · caching-strategy
AI Engineering: prompt-engineering · rag-architecture · agentic-ai-patterns · llm-evaluation · ai-safety-guardrails · llm-observability
Review & Docs: technical-writing
Meta: writing-skills · open-source-project · product-strategy
Unique: cost-aware-routing · design-with-stitch · design-with-pencil
| Tool | Format | Install |
|---|---|---|
| Claude Code | Plugin (native) | /plugin marketplace add kienbui1995/magic-powers |
| Cursor | .mdc rules |
bash scripts/install.sh → select Cursor |
| GitHub Copilot | Agent .md files |
bash scripts/install.sh → select Copilot |
| Aider | CONVENTIONS.md |
bash scripts/install.sh → select Aider |
| Windsurf | .windsurfrules |
bash scripts/install.sh → select Windsurf |
| Gemini CLI | Skills | bash scripts/install.sh → select Gemini |
| Codex | Skills + AGENTS.md |
bash scripts/install.sh → select Codex |
| Kiro | Steering files | bash scripts/install.sh → select Kiro |
| OpenCode | AGENTS.md |
bash scripts/install.sh → select OpenCode |
# Requires STITCH_API_KEY environment variable
node scripts/stitch.mjs generate <projectId> "A dashboard with sidebar"
node scripts/stitch.mjs get-html <projectId> <screenId>See examples/ for real-world scenarios:
- Feature Development — full lifecycle with cost breakdown
- Production Debugging — incident response
- Database Review — schema + security review
- UI Design with Stitch — design-to-code
| Feature | Superpowers | Magic Powers |
|---|---|---|
| Skills | 14 | 43 |
| Agents | 1 | 11 |
| Commands | 0 | 11 |
| Model routing | ❌ | ✅ Opus/Sonnet/Haiku |
| Cost optimization | ❌ | ✅ ~75% reduction |
| Optional hooks | ❌ | ✅ Safety, lint, test |
| Design tools | ❌ | ✅ Google Stitch SDK |
| Multi-tool | ❌ | ✅ 9 tools |
| Built-in/Optional arch | ❌ | ✅ |
See CHANGELOG.md for release history.
See CONTRIBUTING.md for how to add agents and skills.
MIT — see LICENSE.
Inspired by superpowers by Jesse Vincent and agency-agents by Mike Sitarzewski.