English | Tiếng Việt
CodyMaster
"I can't write code. But in 6 months, I shipped 12 real products using AI. CodyMaster is everything I learned — so you don't have to repeat my mistakes." — Tody Le, Head of Product, Creator of CodyMaster
50+ skills. One install. Your AI coding agent becomes a full team.
v7.0.2 — Browse Hybrid Bridge: AI-native browser automation with a11y snapshots, error collection, video recording.
( . \ --- / . )
/ ^ ^ \
( u )
| \ ___ / |
'--w---w--'
Meet Cody 🐹
The Problem
You installed an AI coding agent. It writes code fast. But then:
- It designs differently every time — your brand looks like 3 different companies
- It fixes one thing, breaks five others — you redo the same work over and over
- It forgets everything between sessions — you re-explain your project every morning
- It writes zero tests, zero docs — your codebase becomes a ticking time bomb
- You install 15 skills — none of them talk to each other
"AI gave me 100 hands. Without discipline, those hands created chaos."
The Solution: A Full Senior Team in One Install
CodyMaster isn't a collection of random skills. It's an operating system for your AI agent — 50+ skills that work together like a real team.
When you install CodyMaster, you hire:
| Role | What They Do | Key Skills |
|---|
| Senior Developer | Write tests first, debug systematically, review code | cm-tdd cm-debugging cm-code-review |
| UX Designer | Extract design systems, preview UI before coding | cm-design-system cm-ux-master cm-open-design |
| Product Manager | Plan features, brainstorm ideas, understand users | cm-planning cm-brainstorm-idea |
| DevOps Engineer | Safe deploys, secret scanning, right-account checks | cm-safe-deploy cm-identity-guard |
| Tech Writer | Auto-generate docs, API refs, knowledge base from code | cm-dockit cm-codeintell |
| Automation Lead | Sprint pipelines, autopilot workflows, status tracking | cm-sprint-bus cm-autopilot |
Quick Start — 30 Seconds
Step 1: Install
npm install -g codymaster && cm
A wizard detects your AI tools (Claude Code, Cursor, Gemini, etc.) and installs skills to all of them.
Step 2: Start coding
Open your AI agent and say what you want:
Build a landing page for my coffee shop
Step 3: Watch it work
CodyMaster automatically chains the right skills:
cm-brainstorm-idea → cm-design-system → cm-execution → cm-safe-deploy
That's it. No config files. No setup ritual. Just describe what you want.
Disciplined Delivery by Default
CodyMaster now carries one execution contract from planning through verification:
- Rich plans name exact task files and keep every step inside that task scope; TDD work must record RED before GREEN.
- One approved plan grants scoped execution authorization, so in-scope work continues without repeated confirmation loops.
- Mode B uses three distinct sessions per task: a fresh implementer, a spec reviewer, and then a quality reviewer.
- The shared autonomy policy is validated and synchronized across all 14 supported platform distributions.
cm bench --suite workflow-integration --runs 1 runs from any project directory, including an installed npm package, and writes results locally.
See It Work
Here's a real conversation — a non-technical founder building a booking site:
You: I want to build a booking website for my spa business.
I have an old website at myspa.com but it looks terrible.
AI: [cm-brainstorm-idea] Let me analyze your current site first...
I found 3 key issues: no mobile booking, unclear pricing,
slow load time. Here are 2 approaches:
A) Redesign from scratch (recommended — your brand needs it)
B) Fix the existing site
You: Let's go with A.
AI: [cm-design-system] I'm extracting your brand colors from
myspa.com... Found: gold + navy. Building a design system.
[cm-planning] Here's the plan: booking page, services page,
about page, contact. Estimated: 30 minutes of AI work.
You: Sounds good, do it.
AI: [cm-tdd] Writing tests first...
[cm-execution] Building 4 pages...
[cm-dockit] Generating service descriptions and FAQ docs...
[cm-quality-gate] All tests pass. Performance: 98/100.
[cm-safe-deploy] Deploying to staging...
✅ Done. Preview: https://staging-myspa.pages.dev
You: The booking button is too small on mobile.
AI: [cm-debugging] Found it — padding was 8px, should be 16px.
[cm-code-review] Fix looks good. No side effects.
[cm-safe-deploy] Deployed fix.
✅ Fixed. Try now.
No code written by the human. No bugs introduced by the AI. The system caught the mobile issue before it reached production.
What Makes This Different
| 😵 Random Skills | 🧠 CodyMaster |
|---|
| Integration | Each skill is standalone | 50+ skills that chain and share memory |
| Memory | Forgets everything between sessions | Remembers your project, your style, your mistakes |
| Safety | Deploy and pray | Multi-layer protection: tests → security → staging → production |
| Design | Random UI every time | Extracts your brand, enforces consistency |
| Documentation | "Maybe later" | Auto-generates docs from your code |
| Self-improvement | Static — what you install is what you get | Learns from feedback, gets better over time |
Built For Builders, Not Coders
CodyMaster was built for people who have ideas, not CS degrees.
You're a good fit if you:
- Want to type a message and get a working product back
- Want your AI to learn from its mistakes
- Don't want to manually review every line of AI output
- Want to deploy with confidence, not prayer
You're NOT a good fit if you:
- Enjoy doing the same setup for every project
- Prefer slow, manual deploys with no safety net
- Like re-explaining your codebase every morning
The 10-Second Tour
Instead of listing 50+ skills, here's what CodyMaster does in plain language:
💡 Before You Build
- Thinks before coding — asks questions, challenges your assumptions, saves you from building the wrong thing
- Plans the architecture — diagrams, data flow, edge cases — all before writing code
🎨 While Building
- Enforces your design system — consistent colors, fonts, spacing across every page
- Writes tests first — catches bugs before they exist
- Debugs systematically — traces the root cause, doesn't guess
🚀 Before Shipping
- Scans for secrets — no API keys leaking to GitHub
- Checks the right account — no accidental deploys to the wrong Cloudflare
- Runs quality gates — no deploy without passing tests + performance checks
📈 After Shipping
- Generates documentation — reads your code, writes the docs
- Learns from feedback — gets smarter with every project
- Runs retrospectives — captures what worked, what didn't, improves the process
📖 Full skill reference →
The Design Pipeline
CodyMaster treats design as a first-class discipline — not an afterthought. Your AI doesn't just write code; it enforces visual consistency across every page.
Three ways to get a design system:
| Method | When to Use | How |
|---|
| Extract from URL | You have an existing brand/site | cm-open-design analyzes your site, extracts colors, fonts, spacing |
| Pick from 129 systems | You want a proven aesthetic | Choose from Linear, Stripe, Vercel, Notion, Apple, Tesla, and 123 more |
| Choose a direction | You have no brand yet | Pick from 5 curated styles: Editorial, Modern Minimal, Warm Soft, Tech Utility, Brutalist |
Extract → Tokens → Validate → Build → QA
│ │ │ │ │
│ │ │ │ └─ cm-quality-gate
│ │ │ └─ cm-execution
│ │ └─ cm-ux-master (48 UX laws)
│ └─ cm-design-system (STITCH_TOKENS)
└─ cm-open-design (129 systems + extraction)
📖 Full design pipeline guide →
Real-World Scenarios
| You Say | What Happens |
|---|
| "Fix this bug" | cm-debugging finds root cause → cm-tdd writes test → cm-quality-gate verifies |
| "Build a landing page" | cm-brainstorm-idea → cm-design-system → cm-execution → cm-safe-deploy |
| "Make it look like Stripe" | cm-open-design extracts Stripe tokens → cm-design-system applies → cm-execution builds |
| "Deploy to production" | cm-secret-shield → cm-security-gate → cm-identity-guard → cm-safe-deploy |
| "What does this code do?" | cm-codeintell reads the codebase → cm-dockit generates docs |
| "Add Vietnamese support" | cm-safe-i18n extracts strings → translates → validates → ships |
| "Start a new project" | cm-project-bootstrap scaffolds → cm-planning plans → cm-execution builds |
Install
One command
npm install -g codymaster && cm
What the wizard does
- Detects every AI tool you have (Claude Code, Cursor, Gemini CLI, Codex, OpenCode, Windsurf, Cline, Aider, Continue, Kiro, Amazon Q, Amp, Copilot, Claude Desktop)
- Asks which ones you want skills for (pre-checked for detected tools)
- Installs to each platform's native location
- Done — open your AI agent and start building
After install
cm doctor # Check what's installed
cm status # See your tasks and progress
cm dashboard # Open the visual dashboard
For OpenCode users
OpenCode supports both skills and plugins:
git clone https://github.com/tody-agent/codymaster.git ~/.cody-master
# Skills (auto-load)
ln -s ~/.cody-master/.opencode/skills ~/.opencode/skills
# Plugins (optional — adds custom tools)
mkdir -p ~/.opencode/plugins
ln -s ~/.cody-master/.opencode/plugins/cm-brainstorm-idea.ts ~/.opencode/plugins/
Restart OpenCode to activate. The cm-brainstorm-idea plugin adds a strategic analysis tool with auto-detection for brainstorm keywords.
📖 Full OpenCode installation guide →
For Codex users
Fastest path:
git clone https://github.com/tody-agent/codymaster.git ~/.cody-master
cd ~/.cody-master
npm ci
npm run build:platforms
Then tell Codex:
Fetch and follow instructions from https://raw.githubusercontent.com/tody-agent/codymaster/main/.codex/INSTALL.md
Or use the local Codex skill tree directly at .codex/skills/.
No Node.js?
bash <(curl -fsSL https://raw.githubusercontent.com/tody-agent/codymaster/main/install.sh) --all --profile core
What's in the package
- 50+ skills — ready-to-use SKILL.md files for every major AI coding agent
- CLI (
cm) — install wizard, doctor, dashboard, MCP server, browse daemon
- Skill profiles — curated bundles:
core, growth, full, knowledge
- Multi-platform — installs to Claude Code, Cursor, Gemini, Codex, OpenCode, Windsurf, Cline, Aider, Continue, Kiro, Amazon Q, Amp, Copilot, Claude Desktop
- Zero config — detects your tools, installs skills, you're done
The Dashboard
Visual mission control for your projects:
┌─────────────────────────────────────────────┐
│ 📊 CodyMaster Dashboard │
│ Tasks │ Progress │ Tokens │ Logs │
└─────────────────────────────────────────────┘
cm dashboard start # Start the dashboard
cm dashboard open # Open in browser
Track tasks, monitor progress, see what your AI team is doing — all in one screen.
How It Works Under the Hood
For the curious (skip this if you just want to build):
- Memory System — Your AI remembers your project across sessions. No more re-explaining.
- Skill Chaining — Skills talk to each other. Planning feeds into design, design feeds into code, code feeds into tests.
- Behavioral Discipline — Inspired by Andrej Karpathy's AI coding rules. Your AI thinks before coding, keeps things simple, makes surgical changes.
- Multi-Layer Safety — Tests → security scan → staging → production. Each layer catches what the previous missed.
- Self-Healing — Skills learn from failures and improve over time.
📖 Architecture deep dive →
For Teams & Advanced Users
Multiple AI Agents
Works with 14+ platforms out of the box:
cm install claude-code --profile core
cm install cursor --profile growth
cm install gemini --profile full
MCP Server
Use CodyMaster as an MCP server for Claude Desktop:
npx codymaster mcp-serve --install-claude
Goose Integration
cm mcp-serve --print-config # Paste into Goose config
📖 All integrations →
CLI Commands
| Command | Description |
|---|
cm | Launch interactive wizard (auto-detect AI tools) |
cm doctor | Check installation health across all platforms |
cm status | View current tasks and progress |
cm dashboard | Open visual mission control |
cm install <platform> --profile <name> | Install skills to a specific platform |
cm update --full | Update all skills to latest |
cm upgrade | Upgrade CodyMaster itself |
cm mcp-serve | Run as MCP server for Claude Desktop |
cm browse | Open a URL in the browse daemon |
cm browse screenshot <url> | Capture a screenshot |
cm browse snapshot <url> | Get accessibility tree snapshot |
Skill Profiles
| Profile | Skills | Best For |
|---|
core | 15 | Daily coding — planning, TDD, debug, review, deploy |
growth | 25 | + design system, i18n, content factory |
full | 50+ | Everything — full senior team |
knowledge | 10 | Docs, code intelligence, retros |
cm install claude-code --profile core
cm install cursor --profile growth
cm install gemini --profile full
Troubleshooting
cm: command not found
# Make sure global npm bin is in PATH
npm config get prefix
# Add the output bin/ folder to your PATH
Skills not showing in AI agent
cm doctor # Check what's installed where
cm install --all # Re-install to all detected platforms
Build fails
npm run build # Rebuild TypeScript
npm run test:gate:kit # Run full quality gate
Permission errors (macOS/Linux)
sudo chown -R $(whoami) $(npm config get prefix)/{lib/node_modules,bin,share}
Contributing
- ⭐ Star the repo — helps more builders find this
- Fork → Create
skills/cm-your-skill/SKILL.md
- Submit a Pull Request
CI runs npm run test:gate:kit on every push and PR.
Resources
Who Built This
Tody Le — Head of Product with 10+ years of experience. Can't write code. Used AI to build real products for 6 months straight. Every skill in this kit was born from a real failure that cost real time and real tears.
"50+ skills. Each skill is a lesson. Each lesson is a sleepless night. And now, you don't have to go through those nights."
ISC License — Free to use, modify, and distribute.
Built with ❤️ for the vibe coding community.
"CodyMaster" = "Code Đi" (Vietnamese: "Go code!") — just start building.
