Module 05: Skills & Automation

Module 05: Skills & Automation

Time: 1.5 hours | Complexity: ⭐⭐ Intermediate

Goal

Create reusable skills that give Claude domain-specific knowledge. Package solutions for repeated problems.

---

What You’ll Learn

• What skills are and why they’re powerful
• Creating skills with SKILL.md
• Skill frontmatter and metadata
• Auto-invoking skills
• Building a knowledge base
• Bundling skills with your projects

---

What Are Skills?

A skill is a reusable knowledge module. It teaches Claude how to do something specific.

Example: Testing Skill

Instead of explaining your testing approach every session, you create a skill:

# Testing Best Practices for Our Project

## Framework: Jest

## Style
- Descriptive names: "should validate email with + symbols"
- Arrange-Act-Assert pattern
- Mock external dependencies
- Test behavior, not implementation

## Coverage Target
Minimum 80%

Now, whenever Claude helps with testing, it reads this skill and follows your approach.

Skills vs Agents

Aspect	Skill	Agent
Purpose	Teach knowledge	Execute tasks
Scope	Domain knowledge	Specialized workflow
Persistence	Remembered each session	Called explicitly
Auto-invoke	Yes (optional)	Manual only
Example	”How we test code"	"The test-writer agent”

---

Creating Your First Skill

Skills are markdown files in .claude/skills/.

Basic Structure

---
name: testing-standards
description: Our project's testing practices and conventions
triggers: [test, testing, jest, spec]
auto_invoke: false
keywords: [jest, unit-test, integration-test, mocking]
version: 1.0.0
---

# Testing Standards

## Framework
Jest with @testing-library/react

## File Organization
- Tests live next to source code
- Naming: `[Component].test.tsx`
- Fixtures in `__fixtures__/`
- Mocks in `__mocks__/`

## Test Structure (AAA)
1. **Arrange**: Set up test data
2. **Act**: Call the function/component
3. **Assert**: Check results

## Example

```typescript
describe('validateEmail', () => {
  it('should accept valid email addresses', () => {
    // Arrange
    const email = 'user@example.com';

    // Act
    const result = validateEmail(email);

    // Assert
    expect(result).toBe(true);
  });
});

Coverage Requirements

• Target: 80% minimum
• Critical paths: 100%
• Types of coverage: line, branch, function

Mocking Strategy

• External APIs: use jest.mock()
• Database: use test fixtures
• Timers: use jest.useFakeTimers()

Running Tests

npm test                 # Run all tests
npm test -- --coverage   # With coverage report
npm test -- --watch      # Watch mode

### File Location

my-project/ └── .claude/ └── skills/ └── testing-standards.md

---

## Skill Features

### Triggers

Automatically invoke the skill when Claude sees certain keywords:

```markdown
---
triggers: [test, jest, spec, coverage, mock]
---

If Claude sees “add tests to this function”, it automatically reads the testing skill.

Auto-Invoke

---
auto_invoke: true
---

When true, Claude loads the skill at session start (without you asking). Use for critical rules.

Keywords

Help Claude’s search find the skill:

---
keywords: [testing, jest, unit-test, mocking, assertions]
---

Version

Track skill versions:

---
version: 1.0.0
---

Update when the skill changes significantly.

---

Common Skill Patterns

Pattern 1: Coding Standards

---
name: python-standards
triggers: [python, flask, django]
---

# Python Coding Standards

## Style
- PEP 8 compliance (max 100 chars)
- Type hints on all functions
- Docstrings in Google format

## Testing
- pytest for unit tests
- 80% minimum coverage
- Mock external dependencies

## File Organization
src/
├── models/
├── services/
├── controllers/
└── tests/

## Imports
```python
# ✅ Good: specific imports
from models import User
from services.auth import authenticate

# ❌ Bad: wildcard imports
from models import *

### Pattern 2: Domain Knowledge

```markdown
---
name: payment-processing
description: Payment system rules and edge cases
auto_invoke: true
---

# Payment Processing Rules

## PCI Compliance
- Never log card numbers
- Use tokenization (Stripe)
- Encrypt sensitive data
- Audit all transactions

## Common Issues
1. Partial charges: Retry with exponential backoff
2. Currency conversion: Always round to 2 decimals
3. Timezone handling: Store all times in UTC

## Edge Cases
- Declined cards: Provide clear error message
- Expired cards: Suggest updating payment method
- 3D Secure: Handle verification flow

Pattern 3: Process Documentation

---
name: code-review-checklist
triggers: [review, pull request, pr]
---

# Code Review Checklist

## Before Requesting Review
- [ ] Tests pass locally
- [ ] No console.log statements
- [ ] No secrets in code
- [ ] Commit messages are clear

## Security Checks
- [ ] No SQL injection vulnerabilities
- [ ] No XSS vulnerabilities
- [ ] No exposed API keys
- [ ] Input is validated

## Performance
- [ ] No N+1 queries
- [ ] No infinite loops
- [ ] Load times acceptable

## Testing
- [ ] Unit tests added
- [ ] Integration tests updated
- [ ] Coverage >80%

---

Bundling Skills

You can package multiple related skills together.

Project Skill Bundle

my-project/
└── .claude/
    └── skills/
        ├── testing-standards.md
        ├── api-design.md
        ├── database-patterns.md
        └── security-checklist.md

In CLAUDE.md, reference them:

## Available Skills
Our custom skills are loaded automatically:
- **testing-standards**: How we write tests
- **api-design**: REST API conventions
- **database-patterns**: Common queries and migrations
- **security-checklist**: Security review process

Distributing Skills

To share skills with your team, version control them in git:

git add .claude/skills/
git commit -m "Add testing and API design skills"
git push

Teammates checkout the project and get the skills automatically.

---

Exercise: Create a Domain Skill

Scenario

You’re building an e-commerce site. You want Claude to understand your product data model.

Step 1: Create the Skill

cat > .claude/skills/product-data-model.md << 'EOF'
---
name: product-data-model
description: E-commerce product data structure and rules
triggers: [product, catalog, sku, price, inventory]
auto_invoke: false
version: 1.0.0
---

# Product Data Model

## Core Entities

### Product

{ id: UUID, name: string, slug: string, // URL-friendly description: string, category_id: UUID, created_at: timestamp, updated_at: timestamp }

### SKU (Stock Keeping Unit)

{ id: UUID, product_id: UUID, sku: string, // e.g., “BLUE-XL-001” price: decimal, // Always 2 decimals cost: decimal, inventory: integer, weight: float, // In kg }

### Inventory Rules
- Decrement on order placement
- Increment on return
- Low stock alert: <5 units
- Reorder level: Set per product

## Common Queries

### Get product with all SKUs
```sql
SELECT p.*, s.*
FROM products p
JOIN skus s ON p.id = s.product_id
WHERE p.slug = ?

Check inventory

SELECT sum(inventory) FROM skus WHERE product_id = ?

Edge Cases

1. Out of stock: Return 404 or “unavailable”
2. Variant selection: Show price per SKU
3. Price changes: Update in SKU, not Product EOF

### Step 2: Reference in CLAUDE.md

```markdown
## Skills
- **product-data-model**: Understanding our product structure

Step 3: Use It

In a session:

Add a query to find all products with low inventory (< 5 units)

Claude will:

1. Read the product-data-model skill
2. Understand your schema
3. Write the correct SQL

---

Best Practices

DO

✅ Create skills for things you repeat

✅ Keep skills focused (one domain per skill)

✅ Version control your skills

✅ Include examples in skills

✅ Update skills when requirements change

✅ Share skills with your team

DON’T

❌ Create skills for one-time knowledge (use CLAUDE.md instead)

❌ Make skills too long (>500 lines = break into multiple skills)

❌ Use skills for temporary instructions (use CLAUDE.md or AGENT.md)

❌ Assume skills are comprehensive documentation

---

Skill Lifecycle

1. Create: Identify repeated pattern or domain knowledge
2. Document: Write the skill with examples
3. Test: Use it in a session and verify Claude follows it
4. Refine: Update based on feedback
5. Share: Commit to git for team access
6. Maintain: Update as your practices evolve

---

Validation: You’re Ready If…

✓ You’ve created at least one custom skill

✓ You understand triggers and auto-invoke

✓ You know the difference between skills and agents

✓ You can explain when to use a skill vs CLAUDE.md

✓ Your skill has been tested in a real session

---

What’s Next?

Module 06: Hooks & Events covers:

• Automating responses to system events
• Pre-commit validation
• Post-action notifications
• Building safe automation

This teaches you how to automate repetitive tasks without manual intervention.

---

Completed Module 05? → Ready for Module 06: Hooks & Events