engineer-expertise-extractor（工程师专业知识提取器）

Engineer Expertise Extractor

Extract and document an engineer's coding expertise by analyzing their GitHub contributions, creating a structured knowledge base that captures their coding style, patterns, best practices, and architectural decisions.

What This Skill Does

Researches an engineer's work to create a "digital mentor" by:

• Analyzing Pull Requests - Extract code patterns, review style, decisions
• Extracting Coding Style - Document their preferences and conventions
• Identifying Patterns - Common solutions and approaches they use
• Capturing Best Practices - Their quality standards and guidelines
• Organizing Examples - Real code samples from their work
• Documenting Decisions - Architectural choices and reasoning

Why This Matters

Knowledge Preservation:

• Capture expert knowledge before they leave
• Document tribal knowledge
• Create mentorship materials
• Onboard new engineers faster

Consistency:

• Align team coding standards
• Replicate expert approaches
• Maintain code quality
• Scale expertise across team

Learning:

• Learn from senior engineers
• Understand decision-making
• See real-world patterns
• Improve code quality

How It Works

1. Research Phase

Using GitHub CLI (gh), the skill:

• Fetches engineer's pull requests
• Analyzes code changes
• Reviews their comments and feedback
• Extracts patterns and conventions
• Identifies their expertise areas

2. Analysis Phase

Categorizes findings into:

• Coding Style - Formatting, naming, structure
• Patterns - Common solutions and approaches
• Best Practices - Quality guidelines
• Architecture - Design decisions
• Testing - Testing approaches
• Code Review - Feedback patterns
• Documentation - Doc style and practices

3. Organization Phase

Creates structured folders:

engineer_profiles/
└── [engineer_name]/
    ├── README.md (overview)
    ├── coding_style/
    │   ├── languages/
    │   ├── naming_conventions.md
    │   ├── code_structure.md
    │   └── formatting_preferences.md
    ├── patterns/
    │   ├── common_solutions.md
    │   ├── design_patterns.md
    │   └── code_examples/
    ├── best_practices/
    │   ├── code_quality.md
    │   ├── testing_approach.md
    │   ├── performance.md
    │   └── security.md
    ├── architecture/
    │   ├── design_decisions.md
    │   ├── tech_choices.md
    │   └── trade_offs.md
    ├── code_review/
    │   ├── feedback_style.md
    │   ├── common_suggestions.md
    │   └── review_examples.md
    └── examples/
        ├── by_language/
        ├── by_pattern/
        └── notable_prs/

Output Structure

Engineer Profile README

Contains:

• Engineer overview
• Areas of expertise
• Languages and technologies
• Key contributions
• Coding philosophy
• How to use this profile

Coding Style Documentation

Captures:

• Naming conventions (variables, functions, classes)
• Code structure preferences
• File organization
• Comment style
• Formatting preferences
• Language-specific idioms

Example:

# Coding Style: [Engineer Name]

## Naming Conventions

### Variables
- Use descriptive names: `userAuthentication` not `ua`
- Boolean variables: `isActive`, `hasPermission`, `canEdit`
- Collections: plural names `users`, `items`, `transactions`

### Functions
- Verb-first: `getUserById`, `validateInput`, `calculateTotal`
- Pure functions preferred
- Single responsibility

### Classes
- PascalCase: `UserService`, `PaymentProcessor`
- Interface prefix: `IUserRepository`
- Concrete implementations: `MongoUserRepository`

## Code Structure

### File Organization
- One class per file
- Related functions grouped together
- Tests alongside implementation
- Clear separation of concerns

### Function Length
- Max 20-30 lines preferred
- Extract helper functions
- Single level of abstraction

Patterns Documentation

Captures:

• Recurring solutions
• Design patterns used
• Architectural patterns
• Problem-solving approaches

Example:

# Common Patterns: [Engineer Name]

## Dependency Injection

Used consistently across services:

\`\`\`typescript
// Pattern: Constructor injection
class UserService {
  constructor(
    private readonly userRepo: IUserRepository,
    private readonly logger: ILogger
  ) {}
}
\`\`\`

**Why:** Testability, loose coupling, clear dependencies

## Error Handling

Consistent error handling approach:

\`\`\`typescript
// Pattern: Custom error types + global handler
class ValidationError extends Error {
  constructor(message: string) {
    super(message);
    this.name = 'ValidationError';
  }
}

// Usage
if (!isValid(input)) {
  throw new ValidationError('Invalid input format');
}
\`\`\`

**Why:** Type-safe errors, centralized handling, clear debugging

Best Practices Documentation

Captures:

• Quality standards
• Testing approaches
• Performance guidelines
• Security practices
• Documentation standards

Example:

# Best Practices: [Engineer Name]

## Testing

### Unit Test Structure
- AAA pattern (Arrange, Act, Assert)
- One assertion per test preferred
- Test names describe behavior
- Mock external dependencies

\`\`\`typescript
describe('UserService', () => {
  describe('createUser', () => {
    it('should create user with valid data', async () => {
      // Arrange
      const userData = { email: 'test@example.com', name: 'Test' };
      const mockRepo = createMockRepository();

      // Act
      const result = await userService.createUser(userData);

      // Assert
      expect(result.id).toBeDefined();
      expect(result.email).toBe(userData.email);
    });
  });
});
\`\`\`

### Test Coverage
- Aim for 80%+ coverage
- 100% coverage for critical paths
- Integration tests for APIs
- E2E tests for user flows

## Code Review Standards

### What to Check
- [ ] Tests included and passing
- [ ] No console.logs remaining
- [ ] Error handling present
- [ ] Comments explain "why" not "what"
- [ ] No hardcoded values
- [ ] Security considerations addressed

Architecture Documentation

Captures:

• Design decisions
• Technology choices
• Trade-offs made
• System design approaches

Example:

# Architectural Decisions: [Engineer Name]

## Decision: Microservices vs Monolith

**Context:** Scaling user service
**Decision:** Start monolith, extract services when needed
**Reasoning:**
- Team size: 5 engineers
- Product stage: MVP
- Premature optimization risk
- Easier debugging and deployment

**Trade-offs:**
- Monolith pros: Simpler, faster development
- Monolith cons: Harder to scale later
- Decision: Optimize for current needs, refactor when hitting limits

## Decision: REST vs GraphQL

**Context:** API design for mobile app
**Decision:** REST with versioning
**Reasoning:**
- Team familiar with REST
- Simple use cases
- Caching easier
- Over-fetching not a problem yet

**When to reconsider:** If frontend needs complex queries

Code Review Documentation

Captures:

• Feedback patterns
• Review approach
• Common suggestions
• Communication style

Example:

# Code Review Style: [Engineer Name]

## Review Approach

### Priority Order
1. Security vulnerabilities
2. Logic errors
3. Test coverage
4. Code structure
5. Naming and style

### Feedback Style
- Specific and constructive
- Explains "why" behind suggestions
- Provides examples
- Asks questions to understand reasoning

### Common Suggestions

**Security:**
- "Consider input validation here"
- "This query is vulnerable to SQL injection"
- "Should we rate-limit this endpoint?"

**Performance:**
- "This N+1 query could be optimized with a join"
- "Consider caching this expensive operation"
- "Memoize this pure function"

**Testing:**
- "Can we add a test for the error case?"
- "What happens if the API returns null?"
- "Let's test the boundary conditions"

**Code Quality:**
- "Can we extract this into a helper function?"
- "This function is doing too many things"
- "Consider a more descriptive variable name"

Using This Skill

Extract Engineer Profile

./scripts/extract_engineer.sh [github-username]

Interactive workflow:

1. Enter GitHub username
2. Select repository scope (all/specific org)
3. Choose analysis depth (last N PRs)
4. Specify focus areas (languages, topics)
5. Extract and organize findings

Output: Structured profile in engineer_profiles/[username]/

Analyze Specific Repository

./scripts/analyze_repo.sh [repo-url] [engineer-username]

Focuses analysis on specific repository contributions.

Update Existing Profile

./scripts/update_profile.sh [engineer-username]

Adds new PRs and updates existing profile.

Research Sources

GitHub CLI Queries

Pull Requests:

gh pr list --author [username] --limit 100 --state all
gh pr view [pr-number] --json title,body,files,reviews,comments

Code Changes:

gh pr diff [pr-number]
gh api repos/{owner}/{repo}/pulls/{pr}/files

Reviews:

gh pr view [pr-number] --comments
gh api repos/{owner}/{repo}/pulls/{pr}/reviews

Commits:

gh api search/commits --author [username]

Analysis Techniques

Pattern Recognition:

• Identify recurring code structures
• Extract common solutions
• Detect naming patterns
• Find architectural choices

Style Extraction:

• Analyze formatting consistency
• Extract naming conventions
• Identify comment patterns
• Detect structural preferences

Best Practice Identification:

• Look for testing patterns
• Find error handling approaches
• Identify security practices
• Extract performance optimizations

Use Cases

1. Onboarding New Engineers

Problem: New engineer needs to learn team standards Solution: Provide senior engineer's profile as reference

Benefits:

• Real examples from codebase
• Understand team conventions
• See decision-making process
• Learn best practices

2. Code Review Training

Problem: Teaching good code review skills Solution: Study experienced reviewer's feedback patterns

Benefits:

• Learn what to look for
• Understand feedback style
• See common issues
• Improve review quality

3. Knowledge Transfer

Problem: Senior engineer leaving, knowledge lost Solution: Extract their expertise before departure

Benefits:

• Preserve tribal knowledge
• Document decisions
• Maintain code quality
• Reduce bus factor

4. Establishing Team Standards

Problem: Inconsistent coding styles across team Solution: Extract patterns from best engineers, create standards

Benefits:

• Evidence-based standards
• Real-world examples
• Buy-in from team
• Consistent codebase

5. AI Agent Training

Problem: Agent needs to code like specific engineer Solution: Provide extracted profile to agent

Benefits:

• Match expert's style
• Follow their patterns
• Apply their best practices
• Maintain consistency

Profile Usage by Agents

When an agent has access to an engineer profile, it can:

Code Generation:

• Follow extracted naming conventions
• Use identified patterns
• Apply documented best practices
• Match architectural style

Code Review:

• Provide feedback in engineer's style
• Check for common issues they'd catch
• Apply their quality standards
• Match their priorities

Problem Solving:

• Use their common solutions
• Follow their architectural approach
• Apply their design patterns
• Consider their trade-offs

Example Agent Prompt:

"Using the profile at engineer_profiles/senior_dev/, write a user service
following their coding style, patterns, and best practices. Pay special
attention to their error handling approach and testing standards."

Best Practices

Research Ethics

DO:

• ✅ Get permission before extracting
• ✅ Focus on public contributions
• ✅ Respect privacy
• ✅ Use for learning and improvement

DON'T:

• ❌ Extract without permission
• ❌ Share profiles externally
• ❌ Include sensitive information
• ❌ Use for performance reviews

Profile Maintenance

Regular Updates:

• Refresh every quarter
• Add new significant PRs
• Update with latest patterns
• Archive outdated practices

Quality Control:

• Verify extracted patterns
• Review examples for relevance
• Update documentation
• Remove deprecated practices

Effective Usage

For Learning:

• Study patterns with context
• Understand reasoning behind choices
• Practice applying techniques
• Ask questions when unclear

For Replication:

• Start with style guide
• Reference patterns for similar problems
• Adapt to current context
• Don't blindly copy

Limitations

What This Extracts:

• ✅ Coding style and conventions
• ✅ Common patterns and approaches
• ✅ Best practices and guidelines
• ✅ Architectural decisions
• ✅ Review feedback patterns

What This Doesn't Capture:

• ❌ Real-time problem-solving process
• ❌ Verbal communication style
• ❌ Meeting discussions
• ❌ Design phase thinking
• ❌ Interpersonal mentoring

Future Enhancements

Potential additions:

• Slack message analysis (communication style)
• Design doc extraction (design thinking)
• Meeting notes analysis (decision process)
• Video analysis (pair programming sessions)
• Code metrics tracking (evolution over time)

Example Output

engineer_profiles/
└── senior_dev/
    ├── README.md
    │   # Senior Dev - Staff Engineer
    │   Expertise: TypeScript, Node.js, System Design
    │   Focus: API design, performance optimization
    │
    ├── coding_style/
    │   ├── typescript_style.md
    │   ├── naming_conventions.md
    │   └── code_structure.md
    │
    ├── patterns/
    │   ├── dependency_injection.md
    │   ├── error_handling.md
    │   └── examples/
    │       ├── service_pattern.ts
    │       └── repository_pattern.ts
    │
    ├── best_practices/
    │   ├── testing_strategy.md
    │   ├── code_quality.md
    │   └── performance.md
    │
    ├── architecture/
    │   ├── api_design.md
    │   ├── database_design.md
    │   └── scaling_approach.md
    │
    ├── code_review/
    │   ├── feedback_examples.md
    │   └── review_checklist.md
    │
    └── examples/
        └── notable_prs/
            ├── pr_1234_auth_refactor.md
            └── pr_5678_performance_fix.md

Summary

This skill transforms an engineer's GitHub contributions into a structured, reusable knowledge base. It captures their expertise in a format that:

• Humans can learn from - Clear documentation with examples
• Agents can replicate - Structured patterns and guidelines
• Teams can adopt - Evidence-based best practices
• Organizations can preserve - Knowledge that survives turnover

The goal: Make expertise scalable, learnable, and replicable.

"The best way to learn is from those who have already mastered it."