Buổi 04: CodyMaster Skill System — Master 60+ Skills 🧠

Thành quả: Decision tree: cho bất kỳ tình huống dev nào → biết dùng skill combo nào

---

🎯 Mục Tiêu

1. Nắm vững 6 skill domains và 60+ skills
2. Xây dựng decision tree cho mọi use case
3. Hiểu skill chaining và dependency graph
4. Thực hành invoke skill combo cho 5 scenarios thực tế
5. Master cm-skill-mastery + cm-skill-index

---

📖 Phần 1: 6 Skill Domains Deep Dive

Domain Map

                    ┌─────────────────────┐
                    │   YOUR OBJECTIVE    │
                    └──────────┬──────────┘
                               │
          ┌────────────────────┼────────────────────┐
          │                    │                    │
    ┌─────┴──────┐      ┌─────┴──────┐      ┌─────┴──────┐
    │ 🔧 BUILD  │      │ ⚙️ SHIP    │      │ 🔒 GUARD  │
    │ Engineering│      │ Operations │      │ Security   │
    └─────┬──────┘      └─────┬──────┘      └─────┬──────┘
          │                    │                    │
    ┌─────┴──────┐      ┌─────┴──────┐      ┌─────┴──────┐
    │ 🎨 DESIGN │      │ 📈 GROW    │      │ 🎯 ORCH   │
    │ Product    │      │ Growth     │      │Orchestrate │
    └────────────┘      └────────────┘      └────────────┘

🔧 Engineering Domain

Skill	When	Input	Output
cm-tdd	Viết code mới	Requirement	Failing test → passing code
cm-debugging	Bug/error	Error message, logs	Root cause + fix
cm-quality-gate	Trước deploy	Codebase	Pass/Fail report
cm-test-gate	Setup test infra	Project	4-layer test suite
cm-code-review	PR workflow	Code changes	Review comments
cm-codeintell	Onboard/analyze	Codebase	Skeleton + Architecture
cm-clean-code	Tech debt	Smelly code	Clean refactored code

⚙️ Operations Domain

Skill	When	Input	Output
cm-safe-deploy	Ship code	Built app	Live URL + smoke test
cm-identity-guard	Before push/deploy	Git config	Verified identity
cm-git-worktrees	Parallel features	Branch name	Isolated worktree
cm-terminal	Any command	Shell command	Logged output
cm-safe-i18n	Translate	Source code	Multi-language files
cm-secret-shield	Anytime	Code + env	No leaked secrets

🎨 Product Domain

Skill	When	Input	Output
cm-brainstorm-idea	Complex initiative	Problem statement	2-3 qualified options
cm-planning	Before coding	Requirements	Implementation plan
cm-ux-master	UI/UX design	Feature description	Design guidelines
cm-ui-preview	Visual validation	Design spec	Preview mockups
cm-project-bootstrap	New project	Project description	Full setup
cm-dockit	Documentation	Codebase	Knowledge base

📈 Growth Domain

Skill	When	Input	Output
cm-content-factory	Content marketing	Niche + keywords	Articles + deploy
cm-ads-tracker	Conversion tracking	Tracking IDs	GTM + Pixel setup
cm-cro-methodology	Optimize conversion	Landing page	A/B test plan

🎯 Orchestration Domain

Skill	When	Input	Output
cm-execution	Run plan	Implementation plan	Executed tasks
cm-continuity	Session management	CONTINUITY.md	Preserved context
cm-skill-chain	Multi-skill workflow	Skill sequence	Chained output
cm-start	Full autonomous	Objective	Production code
cm-deep-search	Large codebase	Query	Relevant code
cm-notebooklm	Knowledge sync	Key patterns	Cloud brain

---

📖 Phần 2: Decision Tree

Scenario → Skill Combo

📋 "Tôi cần..."

├── "...build feature mới"
│   └── cm-planning → cm-tdd → cm-execution → cm-quality-gate → cm-safe-deploy
│
├── "...fix bug production"
│   └── cm-debugging → cm-tdd (reproduce) → cm-code-review → cm-safe-deploy
│
├── "...onboard codebase mới"
│   └── cm-codeintell → cm-dockit → (optional: cm-deep-search)
│
├── "...refactor code cũ"
│   └── cm-codeintell → cm-brainstorm-idea → cm-clean-code → cm-tdd → cm-quality-gate
│
├── "...setup project mới"
│   └── cm-identity-guard → cm-project-bootstrap → cm-planning → cm-tdd
│
├── "...deploy lên production"
│   └── cm-identity-guard → cm-quality-gate → cm-safe-deploy
│
├── "...analyze & improve existing product"
│   └── cm-brainstorm-idea → cm-planning → cm-execution
│
├── "...translate to multi-language"
│   └── cm-safe-i18n → cm-quality-gate
│
├── "...create content/marketing"
│   └── cm-content-factory → cm-ads-tracker → cm-cro-methodology
│
└── "...automate workflow"
    └── cm-skill-chain → cm-execution (parallel mode)

---

📖 Phần 3: Skill Chaining

Dependency Graph

cm-identity-guard ─────────────────────────────────────────┐
        │                                                   │
cm-codeintell ──→ cm-brainstorm-idea ──→ cm-planning       │
                                             │              │
                                      cm-tdd ──→ cm-execution
                                             │              │
                                    cm-quality-gate         │
                                             │              │
                                    cm-safe-deploy ────────┘
                                             │
                                    cm-continuity (update session)

5 Built-in Chains

Chain	Steps	Use Case
New Feature	plan → tdd → execute → quality → deploy	Feature dev
Bug Fix	debug → tdd → review → deploy	Production bugs
Refactor	intell → brainstorm → clean → tdd → quality	Tech debt
Onboard	intell → dockit → deep-search → notebooklm	New project
Ship	identity → quality → deploy → continuity	Release

---

📖 Phần 4: Progressive Disclosure

cm-skill-index — 3 Layer Loading

Layer 1 (Always loaded — ~100 tokens):
→ Skill name + trigger keywords
→ "Oh, user said 'debug' → cm-debugging"

Layer 2 (On context match — ~300 tokens):
→ Skill summary + integration points
→ "cm-debugging: 5-phase investigation, works with cm-tdd"

Layer 3 (On execution — full SKILL.md):
→ Complete instructions, templates, examples
→ "Phase 1: Reproduce → Phase 2: Isolate → ..."

Why this matters: Saves 90% tokens when not using a skill.

---

📖 Phần 5: Practical Scenarios

Scenario 1: "Join existing team project"

Day 1:
→ cm-identity-guard: Verify git account
→ cm-codeintell: Generate skeleton index + architecture diagram
→ cm-dockit: Generate knowledge base from code

Day 2:
→ cm-brainstorm-idea: Assess assigned task complexity
→ cm-planning: Create implementation plan
→ cm-tdd: Start coding with tests

Scenario 2: "Legacy codebase, no tests"

Step 1: cm-codeintell → Understand architecture
Step 2: cm-clean-code → Identify code smells
Step 3: cm-test-gate → Setup test infrastructure
Step 4: cm-tdd → Add tests for critical paths
Step 5: cm-clean-code → Safely refactor with test safety net

Scenario 3: "Under pressure, hotfix needed"

Step 1: cm-debugging → Root cause in 15 min
Step 2: cm-tdd → Reproduce bug with failing test
Step 3: Fix → Make test pass (GREEN)
Step 4: cm-safe-deploy → Deploy via hotfix branch
Step 5: cm-continuity → Document the fix

---

🧪 Lab: Skill Decision Exercise

Bài tập: Cho mỗi tình huống, liệt kê skill chain phù hợp

1. "CTO muốn thêm payment integration (Stripe) vào app hiện tại"
2. "User report: Checkout page blank trên Safari mobile"
3. "Startup mới, cần MVP trong 2 tuần"
4. "Translation team yêu cầu thêm tiếng Nhật"
5. "Code base 200 files, không ai hiểu architecture"

---

🎓 Tóm Tắt

Concept	Key Takeaway
6 Domains	Build, Ship, Guard, Design, Grow, Orchestrate
Decision Tree	Situation → Skill combo (never guess)
Chaining	Skills connect: output of one = input of next
Progressive	Load skill details only when needed

⏭️ Buổi tiếp theo

Buổi 05: Codebase Onboarding — Đọc Hiểu Dự Án Trong 1 Giờ 🔍