Advanced

claude --agents '{
  "code-reviewer": {
    "description": "Expert code reviewer. Use proactively after code changes.",
    "prompt": "You are a senior code reviewer. Focus on security, performance, and best practices.",
    "tools": ["Read", "Grep", "Glob"],
    "model": "sonnet"
  },
  "debugger": {
    "description": "Debugging specialist for errors and test failures.",
    "prompt": "You are an expert debugger. Analyze errors, identify root causes, provide fixes.",
    "tools": ["Read", "Bash", "Grep"],
    "model": "sonnet"
  }
}' \
  "Review the API code and debug any test failures"
                                    # => Claude has access to two custom agents
                                    # => Can invoke code-reviewer for reviews
                                    # => Can invoke debugger for test failures
                                    # => Each agent has specialized tools and prompts

# Main Claude session
You: Analyze this codebase for security issues and performance problems

Claude: I'll delegate this to specialized agents:
  1. Invoking security-scanner agent for vulnerability analysis
  2. Invoking performance-analyzer agent for bottleneck detection

# security-scanner agent (runs in isolation)
# => Reads code files
# => Identifies: SQL injection risk in api/users.ts:45
# => Identifies: Missing input validation in api/posts.ts:67
# => Returns report to main session

# performance-analyzer agent (runs in isolation)
# => Analyzes algorithms
# => Identifies: O(n²) loop in utils/search.ts:23
# => Identifies: Unnecessary re-renders in components/List.tsx
# => Returns report to main session

Claude: [Consolidates reports from both agents]
  Security Issues:
    - SQL injection risk (CRITICAL)
    - Missing validation (HIGH)
  Performance Issues:
    - O(n²) complexity (MEDIUM)
    - Unnecessary re-renders (LOW)

# Define MCP servers in config file
cat > mcp-config.json <<EOF
{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "${GITHUB_TOKEN}"
      }
    },
    "database": {
      "command": "node",
      "args": ["./mcp-servers/database-server.js"],
      "env": {
        "DB_CONNECTION_STRING": "${DATABASE_URL}"
      }
    }
  }
}
EOF

# Use MCP servers
claude --mcp-config ./mcp-config.json \
  "Analyze recent GitHub issues and cross-reference with database error logs"
                                    # => Loads github MCP server
                                    # => Loads database MCP server
                                    # => Claude can now:
                                    #    - Query GitHub API via MCP
                                    #    - Query database via MCP
                                    #    - Correlate issues with errors

# Replace entire system prompt (complete control)
claude --system-prompt "You are a TypeScript expert who only writes type-safe code with comprehensive JSDoc comments. Never use 'any' type. Always include error handling." \
  "Create user authentication module"
                                    # => Overrides default Claude behavior
                                    # => Enforces: TypeScript, no 'any', JSDoc, error handling
                                    # => Generated code follows custom rules

# Append to system prompt (keep defaults, add requirements)
claude --append-system-prompt "Always use functional programming patterns. Prefer immutability and pure functions." \
  "Refactor this class-based code to functional style"
                                    # => Keeps default Claude capabilities
                                    # => Adds: FP preference, immutability requirement
                                    # => More flexible than full replacement

# Load prompt from file (version-controlled prompts)
cat > project-prompt.txt <<EOF
Project Conventions:
- Use Prettier for formatting (2 spaces, single quotes)
- Prefix private functions with underscore
- Write tests in /tests directory matching file structure
- Follow Conventional Commits for messages
EOF

claude --append-system-prompt-file ./project-prompt.txt \
  "Add new feature for user notifications"
                                    # => Loads project conventions from file
                                    # => All generated code follows team standards
                                    # => Conventions version-controlled, team-shared

# Allow only read operations (safe for analysis)
claude --tools "Read,Grep,Glob" \
  --dangerously-skip-permissions \
  "Analyze codebase architecture"
                                    # => Claude can Read, Grep, Glob
                                    # => Cannot Write, Edit, Bash (modification tools)
                                    # => Safe for automated analysis

# Disable all tools except specified
claude --tools "Read" \
  "Explain this codebase"
                                    # => Only Read tool available
                                    # => Cannot search, modify, or run commands
                                    # => Maximum safety for read-only operations

# Allowed vs disallowed tools
claude --allowedTools "Bash(git status)" "Bash(git diff)" "Read" \
       --disallowedTools "Bash(git push)" "Bash(rm *)" "Write" \
  "Review changes and suggest improvements"
                                    # => Allowed: git status, git diff, Read
                                    # => Blocked: git push, rm, Write
                                    # => Fine-grained permission control

# No tools (chat-only mode)
claude --tools "" \
  "Explain dependency injection pattern"
                                    # => No tool access
                                    # => Pure conversation mode
                                    # => Cannot interact with files/commands

#!/bin/bash
# multi-agent-workflow.sh - Complex codebase modernization

# Agent 1: Architecture Analysis
echo "Stage 1: Architecture Analysis"
claude -p "analyze codebase architecture and identify legacy patterns" \
  --output-format json > architecture-analysis.json

# Agent 2: Migration Planning (uses Agent 1 results)
echo "Stage 2: Migration Planning"
cat architecture-analysis.json | \
  claude -c -p "create migration plan for identified legacy patterns" \
  --output-format json > migration-plan.json

# Agent 3: Risk Assessment (uses Agent 2 results)
echo "Stage 3: Risk Assessment"
cat migration-plan.json | \
  claude -c -p "assess migration risks and identify blockers" \
  --output-format json > risk-assessment.json

# Agent 4: Prioritization (consolidates all previous)
echo "Stage 4: Prioritization"
jq -s '.' architecture-analysis.json migration-plan.json risk-assessment.json | \
  claude -p "create prioritized execution roadmap balancing impact vs risk" \
  > final-roadmap.md

echo "✅ Multi-agent workflow complete"
cat final-roadmap.md

# Original session: Analyze authentication system
claude -p "analyze authentication system in src/auth/" \
  --session-id "auth-analysis-main" \
  --output-format json > auth-analysis.json

# Fork 1: Experiment with JWT approach
claude -r "auth-analysis-main" --fork-session \
  --session-id "auth-experiment-jwt" \
  -p "design JWT-based authentication implementation" \
  > jwt-design.md

# Fork 2: Experiment with session-based approach
claude -r "auth-analysis-main" --fork-session \
  --session-id "auth-experiment-session" \
  -p "design session-based authentication implementation" \
  > session-design.md

# Fork 3: Experiment with OAuth integration
claude -r "auth-analysis-main" --fork-session \
  --session-id "auth-experiment-oauth" \
  -p "design OAuth 2.0 authentication implementation" \
  > oauth-design.md

# Compare results
echo "=== JWT Approach ===" && cat jwt-design.md
echo "=== Session Approach ===" && cat session-design.md
echo "=== OAuth Approach ===" && cat oauth-design.md

# Original session unchanged
claude -r "auth-analysis-main" \
  "summarize the original auth analysis"

#!/bin/bash
# production-monitor.sh - Continuous monitoring

while true; do
  # Fetch latest logs
  kubectl logs deployment/api --since=5m > /tmp/recent-logs.txt

  # Fetch metrics
  curl -s http://prometheus:9090/api/v1/query?query=api_error_rate > /tmp/metrics.json

  # Analyze with Claude
  ANALYSIS=$(cat /tmp/recent-logs.txt /tmp/metrics.json | \
    claude -p "analyze logs and metrics for anomalies, errors, or performance degradation. Return JSON: {healthy: bool, issues: [], severity: string}" \
    --output-format json \
    --no-session-persistence)

  # Check health
  HEALTHY=$(echo "$ANALYSIS" | jq -r '.healthy')
  SEVERITY=$(echo "$ANALYSIS" | jq -r '.severity')

  if [ "$HEALTHY" = "false" ]; then
    echo "⚠️ ALERT: $SEVERITY severity issues detected"
    echo "$ANALYSIS" | jq '.issues[]'

    # Notify team (Slack, PagerDuty, etc.)
    ./send-alert.sh "$ANALYSIS"

    if [ "$SEVERITY" = "critical" ]; then
      echo "🚨 CRITICAL: Triggering incident response"
      ./trigger-incident.sh
    fi
  else
    echo "✅ System healthy ($(date))"
  fi

  sleep 300  # Check every 5 minutes
done

#!/bin/bash
# robust-claude-automation.sh

MAX_RETRIES=3
RETRY_DELAY=5

function call_claude_with_retry() {
  local prompt="$1"
  local attempt=0

  while [ $attempt -lt $MAX_RETRIES ]; do
    attempt=$((attempt + 1))
    echo "Attempt $attempt of $MAX_RETRIES..."

    # Call Claude with timeout
    if OUTPUT=$(timeout 60 claude -p "$prompt" --output-format json 2>&1); then
      # Success
      echo "$OUTPUT"
      return 0
    else
      EXIT_CODE=$?
      echo "❌ Attempt $attempt failed (exit code: $EXIT_CODE)"

      if [ $attempt -lt $MAX_RETRIES ]; then
        echo "Retrying in $RETRY_DELAY seconds..."
        sleep $RETRY_DELAY
        RETRY_DELAY=$((RETRY_DELAY * 2))  # Exponential backoff
      fi
    fi
  done

  # All retries failed - fallback strategy
  echo "⚠️ All retries failed. Using fallback..."
  echo '{"error": "Claude unavailable", "fallback": true}' > fallback-result.json
  return 1
}

# Use the robust function
if call_claude_with_retry "analyze code quality"; then
  echo "✅ Analysis successful"
else
  echo "❌ Analysis failed after retries, using fallback"
fi

# config/claude-dev.json (development profile)
{
  "model": "haiku",
  "maxBudgetUsd": "1.00",
  "allowedTools": ["Read", "Write", "Edit", "Bash", "Grep", "Glob"],
  "appendSystemPrompt": "Development mode: prioritize speed over thoroughness"
}

# config/claude-staging.json (staging profile)
{
  "model": "sonnet",
  "maxBudgetUsd": "5.00",
  "allowedTools": ["Read", "Grep", "Glob"],
  "appendSystemPrompt": "Staging validation: thorough analysis required"
}

# config/claude-prod.json (production profile)
{
  "model": "sonnet",
  "maxBudgetUsd": "10.00",
  "allowedTools": ["Read", "Grep"],
  "appendSystemPrompt": "Production analysis: maximum safety, read-only operations"
}

# Usage with environment selection
ENV="${DEPLOY_ENV:-dev}"
CONFIG_FILE="config/claude-${ENV}.json"

# Load config and run Claude
MODEL=$(jq -r '.model' "$CONFIG_FILE")
BUDGET=$(jq -r '.maxBudgetUsd' "$CONFIG_FILE")
TOOLS=$(jq -r '.allowedTools | join(",")' "$CONFIG_FILE")
PROMPT=$(jq -r '.appendSystemPrompt' "$CONFIG_FILE")

claude --model "$MODEL" \
       --max-budget-usd "$BUDGET" \
       --tools "$TOOLS" \
       --append-system-prompt "$PROMPT" \
       -p "analyze security vulnerabilities"

# Output depends on environment:
# - DEV: Fast analysis (haiku), can modify code ($1 budget)
# - STAGING: Thorough analysis (sonnet), read-only ($5 budget)
# - PROD: Maximum analysis (sonnet), strictest read-only ($10 budget)

FROM node:20
# => Base image includes build tools (700MB+)
WORKDIR /app
COPY package*.json ./
# => Copies package files
RUN npm install
# => Installs ALL dependencies (dev + production)
COPY . .
# => Copies source code
RUN npm run build
# => Builds application
CMD ["node", "dist/server.js"]
# => Final image size: 1.2GB (includes build tools, dev deps)

# Stage 1: Build stage
FROM node:20 AS builder
# => Build stage uses full Node image
WORKDIR /app
COPY package*.json ./
# => Copies package files
RUN npm ci --only=production
# => Installs production dependencies only
COPY . .
# => Copies source code
RUN npm run build
# => Builds application

# Stage 2: Production stage
FROM node:20-alpine
# => Alpine Linux base (much smaller, 40MB vs 700MB)
WORKDIR /app
COPY --from=builder /app/dist ./dist
# => Copies ONLY built files from build stage
COPY --from=builder /app/node_modules ./node_modules
# => Copies ONLY production dependencies
COPY package*.json ./
# => Copies package metadata
USER node
# => Runs as non-root user (security)
CMD ["node", "dist/server.js"]
# => Final image size: 150MB (75% reduction)

You: Optimize the Dockerfile for production using multi-stage builds
                                    # => Claude reads current Dockerfile
                                    # => Identifies inefficiencies:
                                    # =>   - Single stage includes build tools (unnecessary in prod)
                                    # =>   - Using full node:20 image (700MB)
                                    # =>   - Installing devDependencies in production
                                    # => Generates multi-stage Dockerfile:
                                    # =>   - Stage 1: Builder (full Node, build tools)
                                    # =>   - Stage 2: Production (Alpine, minimal)
                                    # => Estimates size reduction: 1.2GB → 150MB
                                    # => Adds .dockerignore for faster builds:
                                    # =>   node_modules, .git, *.md, tests

You: Generate Kubernetes manifests for deploying this Node.js API with 3 replicas, health checks, and horizontal autoscaling
                                    # => Claude analyzes application:
                                    # =>   - Detects: Express API on port 3000
                                    # =>   - Needs: Database connection (from env vars)
                                    # =>   - Requires: Secrets for JWT keys
                                    # => Generates manifests:
                                    # => 1. k8s/deployment.yaml
                                    # =>    apiVersion: apps/v1
                                    # =>    kind: Deployment
                                    # =>    metadata:
                                    # =>      name: api-server
                                    # =>    spec:
                                    # =>      replicas: 3
                                    # =>      selector:
                                    # =>        matchLabels:
                                    # =>          app: api-server
                                    # =>      template:
                                    # =>        spec:
                                    # =>          containers:
                                    # =>          - name: api
                                    # =>            image: company/api:latest
                                    # =>            ports:
                                    # =>            - containerPort: 3000
                                    # =>            livenessProbe:           # => Health check
                                    # =>              httpGet:
                                    # =>                path: /health
                                    # =>                port: 3000
                                    # =>              initialDelaySeconds: 30
                                    # =>            readinessProbe:          # => Ready check
                                    # =>              httpGet:
                                    # =>                path: /ready
                                    # =>                port: 3000
                                    # =>            resources:
                                    # =>              requests:
                                    # =>                memory: "256Mi"
                                    # =>                cpu: "250m"
                                    # =>              limits:
                                    # =>                memory: "512Mi"
                                    # =>                cpu: "500m"
                                    # =>            envFrom:
                                    # =>            - configMapRef:
                                    # =>                name: api-config
                                    # =>            - secretRef:
                                    # =>                name: api-secrets
                                    # =>
                                    # => 2. k8s/service.yaml (ClusterIP service)
                                    # => 3. k8s/hpa.yaml (horizontal autoscaler, 2-10 replicas)
                                    # => 4. k8s/configmap.yaml (non-secret config)
                                    # => 5. k8s/ingress.yaml (external access)

You: Analyze .github/workflows/ci.yml and suggest optimizations to reduce build time
                                    # => Claude reads CI workflow
                                    # => Current pipeline (sequential):
                                    # =>   - Checkout (10s)
                                    # =>   - Install deps (120s)
                                    # =>   - Lint (45s)
                                    # =>   - Unit tests (90s)
                                    # =>   - Integration tests (180s)
                                    # =>   - Build (60s)
                                    # =>   Total: 505s (~8.5 minutes)
                                    # =>
                                    # => Identifies optimizations:
                                    # => 1. Parallelize independent jobs
                                    # =>    - Lint + Unit tests can run parallel (not sequential)
                                    # => 2. Cache node_modules
                                    # =>    - 120s install → 15s cache restore
                                    # => 3. Run integration tests only on main/PR
                                    # =>    - Skip on draft PRs (saves 180s)
                                    # => 4. Use smaller runner
                                    # =>    - ubuntu-latest-4core instead of 2core
                                    # =>
                                    # => Generates optimized workflow:
                                    # =>   jobs:
                                    # =>     setup:
                                    # =>       - uses: actions/cache@v3  # => Caches node_modules
                                    # =>         with:
                                    # =>           path: node_modules
                                    # =>           key: ${{ runner.os }}-node-${{ hashFiles('package-lock.json') }}
                                    # =>     lint:
                                    # =>       needs: setup
                                    # =>       runs-on: ubuntu-latest
                                    # =>       - run: npm run lint     # => Parallel with tests
                                    # =>     test:
                                    # =>       needs: setup
                                    # =>       runs-on: ubuntu-latest-4core  # => Faster runner
                                    # =>       - run: npm test         # => Parallel with lint
                                    # =>     integration:
                                    # =>       if: github.event.pull_request.draft == false
                                    # =>       - run: npm run test:integration
                                    # =>
                                    # => Estimated time: 180s (~3 minutes, 64% reduction)

You: Generate Terraform config for AWS infrastructure: VPC with public/private subnets, RDS PostgreSQL, and ECS Fargate cluster
                                    # => Claude generates Terraform modules:
                                    # => 1. modules/vpc/main.tf
                                    # =>    resource "aws_vpc" "main" {
                                    # =>      cidr_block = var.vpc_cidr
                                    # =>      enable_dns_hostnames = true  # => Enables DNS resolution
                                    # =>      tags = { Name = "${var.project}-vpc" }
                                    # =>    }
                                    # =>    resource "aws_subnet" "public" {
                                    # =>      count = 2                    # => 2 AZs for HA
                                    # =>      vpc_id = aws_vpc.main.id
                                    # =>      cidr_block = cidrsubnet(var.vpc_cidr, 4, count.index)
                                    # =>      availability_zone = data.aws_availability_zones.available.names[count.index]
                                    # =>      map_public_ip_on_launch = true  # => Public subnets
                                    # =>    }
                                    # =>    resource "aws_subnet" "private" {
                                    # =>      count = 2                    # => Private subnets for DB
                                    # =>      vpc_id = aws_vpc.main.id
                                    # =>      cidr_block = cidrsubnet(var.vpc_cidr, 4, count.index + 2)
                                    # =>      availability_zone = data.aws_availability_zones.available.names[count.index]
                                    # =>    }
                                    # =>
                                    # => 2. modules/rds/main.tf
                                    # =>    resource "aws_db_instance" "postgres" {
                                    # =>      engine = "postgres"
                                    # =>      instance_class = var.instance_class
                                    # =>      allocated_storage = 20
                                    # =>      storage_encrypted = true     # => Encryption at rest
                                    # =>      multi_az = true              # => High availability
                                    # =>      db_subnet_group_name = aws_db_subnet_group.main.name
                                    # =>      vpc_security_group_ids = [aws_security_group.db.id]
                                    # =>      backup_retention_period = 7  # => 7-day backups
                                    # =>      skip_final_snapshot = false  # => Safety for production
                                    # =>    }
                                    # =>
                                    # => 3. modules/ecs/main.tf (Fargate cluster)
                                    # => 4. main.tf (root module composing all)
                                    # => 5. variables.tf (parameterized config)
                                    # => 6. outputs.tf (exported values)

You: Generate property-based tests for the sortArray function using fast-check
                                    # => Claude reads sortArray implementation
                                    # => Identifies properties to test:
                                    # =>   1. Idempotence: sort(sort(arr)) === sort(arr)
                                    # =>   2. Length preservation: sort(arr).length === arr.length
                                    # =>   3. Element preservation: sorted array has same elements
                                    # =>   4. Order invariant: sorted[i] <= sorted[i+1]
                                    # => Generates property tests:
                                    # => import fc from 'fast-check';
                                    # => import { sortArray } from './sort';
                                    # =>
                                    # => describe('sortArray property tests', () => {
                                    # =>   test('idempotence: sorting twice equals sorting once', () => {
                                    # =>     fc.assert(
                                    # =>       fc.property(
                                    # =>         fc.array(fc.integer()),  # => Generates random int arrays
                                    # =>         (arr) => {
                                    # =>           const once = sortArray(arr);
                                    # =>           const twice = sortArray(once);
                                    # =>           // => Sorting sorted array doesn't change it
                                    # =>           expect(twice).toEqual(once);
                                    # =>         }
                                    # =>       )
                                    # =>     );
                                    # =>   });
                                    # =>
                                    # =>   test('length preservation', () => {
                                    # =>     fc.assert(
                                    # =>       fc.property(
                                    # =>         fc.array(fc.integer()),
                                    # =>         (arr) => {
                                    # =>           const sorted = sortArray(arr);
                                    # =>           // => Sorting doesn't add/remove elements
                                    # =>           expect(sorted.length).toBe(arr.length);
                                    # =>         }
                                    # =>       )
                                    # =>     );
                                    # =>   });
                                    # =>
                                    # =>   test('ordered output', () => {
                                    # =>     fc.assert(
                                    # =>       fc.property(
                                    # =>         fc.array(fc.integer()),
                                    # =>         (arr) => {
                                    # =>           const sorted = sortArray(arr);
                                    # =>           // => Each element <= next element
                                    # =>           for (let i = 0; i < sorted.length - 1; i++) {
                                    # =>             expect(sorted[i]).toBeLessThanOrEqual(sorted[i + 1]);
                                    # =>           }
                                    # =>         }
                                    # =>       )
                                    # =>     );
                                    # =>   });
                                    # => });
                                    # => Runs 100 random inputs per property automatically

# Create custom agent config in .claude/agents/api-developer.md
You: Create a custom agent for API development that knows our conventions
                                    # => Claude generates agent config:
                                    # => ---
                                    # => name: api-developer
                                    # => description: API development agent with project conventions
                                    # => tools:
                                    # =>   - Read
                                    # =>   - Write
                                    # =>   - Edit
                                    # =>   - Grep
                                    # =>   - Bash
                                    # => model: sonnet
                                    # => skills:
                                    # =>   - docs__creating-accessible-diagrams
                                    # => ---
                                    # =>
                                    # => # API Developer Agent
                                    # =>
                                    # => ## Project Conventions
                                    # =>
                                    # => When creating API endpoints:
                                    # => 1. Route handlers: src/api/[resource].ts
                                    # => 2. Validation: Zod schemas in same file
                                    # => 3. Business logic: src/services/[Resource]Service.ts
                                    # => 4. Data access: src/database/repositories/[Resource]Repository.ts
                                    # => 5. Tests: src/api/[resource].test.ts
                                    # =>
                                    # => ## Error Handling
                                    # =>
                                    # => Use ApiError class:
                                    # => - 400: ValidationError
                                    # => - 404: NotFoundError
                                    # => - 409: ConflictError
                                    # => - 500: InternalError
                                    # =>
                                    # => ## Authentication
                                    # =>
                                    # => Protected routes use authMiddleware.
                                    # => User info available in req.user.

You: Analyze code metrics for this pull request and check if it meets quality gates
                                    # => Claude runs multiple checks:
                                    # => 1. Complexity analysis
                                    # =>    - Uses Bash: npx eslint --format json
                                    # =>    - Finds: 3 functions with cyclomatic complexity > 10
                                    # =>      - src/services/order.ts:45 - complexity 15 (threshold: 10)
                                    # =>      - src/utils/pricing.ts:89 - complexity 12 (threshold: 10)
                                    # =>      - src/api/checkout.ts:120 - complexity 11 (threshold: 10)
                                    # => 2. Code duplication
                                    # =>    - Uses jscpd (copy-paste detector)
                                    # =>    - Finds: 2.3% duplication (threshold: 5%) ✓
                                    # => 3. Test coverage
                                    # =>    - Uses Bash: npm run test:coverage
                                    # =>    - Current: 78% (threshold: 80%) ✗
                                    # =>    - Missing coverage:
                                    # =>      - src/services/order.ts (65% coverage)
                                    # => 4. Type safety (TypeScript)
                                    # =>    - Uses: npx tsc --noEmit
                                    # =>    - Errors: 0 ✓
                                    # => 5. Dependency vulnerabilities
                                    # =>    - Uses: npm audit
                                    # =>    - Critical: 0, High: 0, Moderate: 2 (acceptable)
                                    # =>
                                    # => Quality Gate Result: ✗ FAILED
                                    # => - Complexity violations: 3 functions
                                    # => - Coverage below threshold: 78% < 80%
                                    # =>
                                    # => Recommendations:
                                    # => 1. Refactor high-complexity functions (split into smaller)
                                    # => 2. Add tests to src/services/order.ts to reach 80% coverage

You: Run a security audit on the codebase
                                    # => Claude performs multiple scans:
                                    # => 1. SQL Injection Detection
                                    # =>    - Grep pattern: String concatenation in SQL queries
                                    # =>    - Searches: db.query.*\+|`\${.*}`
                                    # =>    - Finds: 2 potential vulnerabilities
                                    # =>      - src/api/users.ts:45
                                    # =>        db.query(`SELECT * FROM users WHERE id = ${userId}`)
                                    # =>        RISK: SQL injection if userId not sanitized
                                    # =>      - src/api/posts.ts:89
                                    # =>        db.query("DELETE FROM posts WHERE id = " + postId)
                                    # =>        RISK: SQL injection possible
                                    # => 2. XSS Vulnerability Detection
                                    # =>    - Pattern: dangerouslySetInnerHTML, innerHTML
                                    # =>    - Finds: 1 instance
                                    # =>      - src/components/Article.tsx:34
                                    # =>        dangerouslySetInnerHTML={{ __html: article.content }}
                                    # =>        RISK: XSS if content not sanitized
                                    # => 3. Hardcoded Secrets Detection
                                    # =>    - Pattern: api_key, secret, password in code
                                    # =>    - Finds: 1 violation
                                    # =>      - src/config/aws.ts:5
                                    # =>        const AWS_SECRET = "sk_live_12345abcde"
                                    # =>        CRITICAL: Hardcoded AWS secret key
                                    # => 4. Dependency Vulnerabilities
                                    # =>    - Runs: npm audit
                                    # =>    - Finds: 3 high-severity vulnerabilities
                                    # =>      - lodash 4.17.19 (prototype pollution)
                                    # =>      - axios 0.21.1 (SSRF vulnerability)
                                    # =>    - Suggests: npm audit fix --force
                                    # =>
                                    # => Security Report:
                                    # => CRITICAL: 1 (hardcoded secret)
                                    # => HIGH: 5 (2 SQL injection, 1 XSS, 2 dependency)
                                    # => MEDIUM: 0
                                    # =>
                                    # => Immediate actions required:
                                    # => 1. Remove hardcoded AWS secret, use env vars
                                    # => 2. Use parameterized queries for SQL
                                    # => 3. Sanitize HTML before dangerouslySetInnerHTML
                                    # => 4. Update vulnerable dependencies

You: Check all npm dependencies for license compliance. Flag any GPL or AGPL licenses.
                                    # => Claude uses license-checker tool
                                    # => Runs: npx license-checker --json
                                    # => Analyzes 247 dependencies
                                    # => Categorizes by license:
                                    # =>   MIT: 189 packages ✓ (compatible)
                                    # =>   Apache-2.0: 32 packages ✓ (compatible)
                                    # =>   ISC: 18 packages ✓ (compatible)
                                    # =>   BSD-3-Clause: 5 packages ✓ (compatible)
                                    # =>   GPL-3.0: 2 packages ✗ (restricted)
                                    # =>   AGPL-3.0: 1 package ✗ (restricted)
                                    # =>
                                    # => License Violations Found:
                                    # => 1. package-name@1.2.3 (GPL-3.0)
                                    # =>    - Requires derivative works to be GPL
                                    # =>    - Incompatible with proprietary software
                                    # =>    - Recommendation: Find MIT alternative
                                    # => 2. another-pkg@4.5.6 (GPL-3.0)
                                    # =>    - Same issue as above
                                    # => 3. agpl-package@0.0.1 (AGPL-3.0)
                                    # =>    - Network copyleft (affects SaaS)
                                    # =>    - CRITICAL: Must remove or open-source entire app
                                    # =>
                                    # => Suggests alternatives:
                                    # =>   package-name → alternative-mit-pkg
                                    # =>   another-pkg → drop (unused in code)
                                    # =>   agpl-package → reimplement functionality

You: Generate an Architecture Decision Record for our choice of PostgreSQL over MongoDB
                                    # => Claude generates ADR:
                                    # => # ADR-001: Use PostgreSQL for Primary Database
                                    # =>
                                    # => Date: 2026-02-02
                                    # => Status: Accepted
                                    # =>
                                    # => ## Context
                                    # =>
                                    # => We need a primary database for the e-commerce platform.
                                    # => Requirements:
                                    # => - ACID transactions for order processing
                                    # => - Complex queries joining users, orders, products
                                    # => - Strong consistency for inventory management
                                    # => - Mature ecosystem and community support
                                    # =>
                                    # => ## Decision
                                    # =>
                                    # => We will use PostgreSQL as our primary database.
                                    # =>
                                    # => ## Consequences
                                    # =>
                                    # => ### Positive
                                    # => - ACID transactions prevent double-charging
                                    # => - SQL joins enable complex reporting queries
                                    # => - Strong typing prevents data corruption
                                    # => - Mature tooling (pg_dump, pg_stat_statements)
                                    # => - Team has PostgreSQL expertise
                                    # =>
                                    # => ### Negative
                                    # => - Vertical scaling limits (mitigated by read replicas)
                                    # => - Schema migrations require downtime (mitigated by online DDL)
                                    # => - Less flexible schema than document stores
                                    # =>
                                    # => ### Neutral
                                    # => - Requires SQL knowledge (team already has)
                                    # => - Need ORM or query builder (using Prisma)
                                    # =>
                                    # => ## Alternatives Considered
                                    # =>
                                    # => ### MongoDB
                                    # => - Pros: Flexible schema, horizontal scaling
                                    # => - Cons: No transactions (at the time), eventual consistency risks
                                    # => - Rejected: ACID transactions required for e-commerce
                                    # =>
                                    # => ### MySQL
                                    # => - Pros: Mature, familiar
                                    # => - Cons: Less feature-rich than PostgreSQL (no JSON, arrays)
                                    # => - Rejected: PostgreSQL JSON support useful for product attributes
                                    # =>
                                    # => ## Related Decisions
                                    # => - ADR-002: Use Prisma ORM
                                    # => - ADR-005: Redis for caching

You: Analyze the Node.js profiling data and identify performance bottlenecks
                                    # => Claude reads profile output (--prof or clinic)
                                    # => Uses: node --prof app.js → isolate-*.log
                                    # => Runs: node --prof-process isolate-*.log
                                    # => Analyzes CPU profile:
                                    # =>   Total samples: 10,000
                                    # =>   Top functions by CPU time:
                                    # =>   1. JSON.parse() - 2,345 samples (23.45%)
                                    # =>      - Called from: src/middleware/bodyParser.ts
                                    # =>      - Issue: Parsing large request bodies
                                    # =>   2. bcrypt.hash() - 1,890 samples (18.90%)
                                    # =>      - Called from: src/services/auth.ts
                                    # =>      - Issue: High bcrypt rounds (12, typical: 10)
                                    # =>   3. db.query() - 1,234 samples (12.34%)
                                    # =>      - Called from: src/api/products.ts
                                    # =>      - Issue: N+1 query pattern (loading related data)
                                    # => Identifies optimizations:
                                    # =>   1. JSON.parse bottleneck:
                                    # =>      - Add request body size limit (100KB)
                                    # =>      - Use streaming parser for large payloads
                                    # =>   2. bcrypt bottleneck:
                                    # =>      - Reduce rounds from 12 to 10 (still secure)
                                    # =>      - Move hashing to background job for registration
                                    # =>   3. Database N+1:
                                    # =>      - Use JOIN instead of multiple queries
                                    # =>      - Add DataLoader for batching
                                    # => Estimates improvements: 45% faster response times

You: Design an API versioning strategy for our microservices that supports gradual migration
                                    # => Claude proposes versioning approaches:
                                    # => Option 1: URL Versioning
                                    # =>   - /api/v1/users, /api/v2/users
                                    # =>   - Pros: Clear, cache-friendly, widely understood
                                    # =>   - Cons: URL proliferation, requires routing changes
                                    # => Option 2: Header Versioning
                                    # =>   - Accept: application/vnd.company.v2+json
                                    # =>   - Pros: Clean URLs, follows REST principles
                                    # =>   - Cons: Less discoverable, harder to test manually
                                    # => Option 3: Query Parameter
                                    # =>   - /api/users?version=2
                                    # =>   - Pros: Easy to implement
                                    # =>   - Cons: Breaks REST semantics, cache issues
                                    # =>
                                    # => Recommends: URL versioning for microservices
                                    # => Rationale:
                                    # =>   - Clear for service mesh routing
                                    # =>   - Compatible with API gateways
                                    # =>   - Easy for clients to discover versions
                                    # =>
                                    # => Generates implementation:
                                    # =>   1. Router middleware: src/middleware/versioning.ts
                                    # =>      - Extracts version from URL path
                                    # =>      - Routes to version-specific handlers
                                    # =>   2. Version handlers: src/api/v1/, src/api/v2/
                                    # =>      - Separate handlers per version
                                    # =>      - Can share common logic via services
                                    # =>   3. Deprecation headers:
                                    # =>      - X-API-Deprecation: "v1 deprecated 2026-12-31"
                                    # =>      - Sunset: Sat, 31 Dec 2026 23:59:59 GMT
                                    # =>   4. Migration guide generator
                                    # =>      - Documents breaking changes v1 → v2

You: Create a custom ESLint rule that prevents importing from infrastructure layer in domain layer
                                    # => Claude generates ESLint plugin:
                                    # => // eslint-plugin-architecture/no-infrastructure-in-domain.js
                                    # => module.exports = {
                                    # =>   meta: {
                                    # =>     type: 'problem',
                                    # =>     docs: {
                                    # =>       description: 'Prevent domain layer from importing infrastructure',
                                    # =>       category: 'Architecture',
                                    # =>     },
                                    # =>     messages: {
                                    # =>       violation: 'Domain layer cannot import from infrastructure layer',
                                    # =>     },
                                    # =>   },
                                    # =>   create(context) {
                                    # =>     return {
                                    # =>       ImportDeclaration(node) {
                                    # =>         const filename = context.getFilename();
                                    # =>         const importPath = node.source.value;
                                    # =>
                                    # =>         // Check if file is in domain layer
                                    # =>         const isDomainFile = filename.includes('/src/domain/');
                                    # =>         // Check if importing from infrastructure
                                    # =>         const isInfraImport = importPath.includes('infrastructure');
                                    # =>
                                    # =>         if (isDomainFile && isInfraImport) {
                                    # =>           context.report({
                                    # =>             node,
                                    # =>             messageId: 'violation',
                                    # =>           });
                                    # =>         }
                                    # =>       },
                                    # =>     };
                                    # =>   },
                                    # => };
                                    # =>
                                    # => Creates package structure:
                                    # => eslint-plugin-architecture/
                                    # =>   ├── index.js (exports rules)
                                    # =>   ├── rules/
                                    # =>   │   └── no-infrastructure-in-domain.js
                                    # =>   ├── tests/
                                    # =>   │   └── no-infrastructure-in-domain.test.js
                                    # =>   └── package.json
                                    # =>
                                    # => Updates .eslintrc.js:
                                    # => plugins: ['architecture'],
                                    # => rules: {
                                    # =>   'architecture/no-infrastructure-in-domain': 'error',
                                    # => }

You: Generate release notes for version 2.1.0 from commits since v2.0.0
                                    # => Claude runs: git log v2.0.0..HEAD --oneline
                                    # => Analyzes 47 commits
                                    # => Categorizes by conventional commit type:
                                    # =>   feat: 12 commits (new features)
                                    # =>   fix: 23 commits (bug fixes)
                                    # =>   docs: 5 commits (documentation)
                                    # =>   refactor: 4 commits (internal changes)
                                    # =>   perf: 2 commits (performance)
                                    # =>   BREAKING CHANGE: 1 commit
                                    # =>
                                    # => Reads PR descriptions for each commit
                                    # => Generates release notes:
                                    # => # Release v2.1.0 (2026-02-02)
                                    # =>
                                    # => ## ⚠ Breaking Changes
                                    # => - **API**: Changed POST /api/users to require email verification
                                    # =>   - Migration: Update clients to handle 201 response with `verified: false`
                                    # =>   - Clients must call POST /api/users/verify with email token
                                    # =>
                                    # => ## 🚀 Features
                                    # => - Add user profile customization (#234)
                                    # => - Support OAuth2 login with Google and GitHub (#245)
                                    # => - Implement dark mode toggle (#256)
                                    # => - Add export to CSV for reports (#267)
                                    # => - (8 more features...)
                                    # =>
                                    # => ## 🐛 Bug Fixes
                                    # => - Fix memory leak in WebSocket connections (#289)
                                    # => - Resolve race condition in order processing (#291)
                                    # => - Correct timezone handling in date picker (#298)
                                    # => - (20 more fixes...)
                                    # =>
                                    # => ## ⚡ Performance
                                    # => - Optimize database queries with indexes (40% faster)
                                    # => - Reduce bundle size by 25% with code splitting
                                    # =>
                                    # => ## 📚 Documentation
                                    # => - Updated API reference
                                    # => - Added authentication tutorial
                                    # => - (3 more docs updates...)
                                    # =>
                                    # => ## Contributors
                                    # => Thanks to @alice, @bob, @charlie for contributions!

You: Configure Claude to understand our Nx monorepo structure with 15 packages
                                    # => Claude analyzes workspace:
                                    # => Reads: nx.json, workspace.json
                                    # => Identifies packages:
                                    # =>   Libraries:
                                    # =>     - @company/core (shared utilities)
                                    # =>     - @company/auth (authentication)
                                    # =>     - @company/ui (React components)
                                    # =>   Apps:
                                    # =>     - @company/web (Next.js app)
                                    # =>     - @company/api (Express server)
                                    # =>     - @company/mobile (React Native)
                                    # =>
                                    # => Creates .claude/agents/monorepo-dev.md:
                                    # => ---
                                    # => name: monorepo-dev
                                    # => description: Nx monorepo development agent
                                    # => ---
                                    # =>
                                    # => # Monorepo Development Agent
                                    # =>
                                    # => ## Workspace Structure
                                    # =>
                                    # => This is an Nx monorepo with:
                                    # => - 15 packages (12 libraries, 3 apps)
                                    # => - Import aliases: @company/[package-name]
                                    # => - Build system: Nx with caching
                                    # =>
                                    # => ## Package Conventions
                                    # =>
                                    # => ### Libraries (packages/*)
                                    # => - No dependencies on apps
                                    # => - Can depend on other libraries
                                    # => - Export via index.ts (public API)
                                    # => - Tests: *.test.ts (same directory)
                                    # =>
                                    # => ### Apps (apps/*)
                                    # => - Can import any library
                                    # => - Never imported by libraries
                                    # => - Each app is independently deployable
                                    # =>
                                    # => ## Nx Commands
                                    # =>
                                    # => When running tasks:
                                    # => - Build single: nx build [package]
                                    # => - Test single: nx test [package]
                                    # => - Build affected: nx affected:build
                                    # => - Lint all: nx run-many --target=lint --all
                                    # =>
                                    # => ## Dependency Rules
                                    # =>
                                    # => - Libraries CANNOT import from apps
                                    # => - Apps CAN import from libraries
                                    # => - Circular dependencies are forbidden
                                    # =>
                                    # => When creating new packages:
                                    # => 1. Use: nx generate @nrwl/[type]:library [name]
                                    # => 2. Update tsconfig paths for @company/[name]
                                    # => 3. Add to nx.json if needed