configure-api-gateway
について
このスキルは、マイクロサービス全体のトラフィック管理、セキュリティ、ルーティングを統合するAPIゲートウェイとして、KongまたはTraefikを導入・設定します。認証、レート制限、リクエスト変換、負荷分散といった重要な機能を扱います。複数のバックエンドサービスに対する一元化されたエントリーポイントが必要な場合や、一貫したAPIポリシーと分析を求めるときにご利用ください。
クイックインストール
Claude Code
推奨npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/configure-api-gatewayこのコマンドをClaude Codeにコピー&ペーストしてスキルをインストールします
ドキュメント
Configure API Gateway
Deploy + configure API gateway → centralized API traffic mgmt + policy enforcement.
Use When
- Multi backend services need unified API endpoint + consistent policies
- Need centralized auth for API access
- Need rate limiting + quota mgmt across APIs
- Want transform reqs/res w/o modifying backend services
- Impl API versioning + deprecation strategies
- Need detailed API analytics + monitoring
- Need service discovery + load balancing for microservices
In
- Required: Kubernetes cluster or Docker env
- Required: Choice of API gateway (Kong or Traefik)
- Required: Backend service endpoints to proxy
- Optional: Auth provider (OAuth2, OIDC, API keys)
- Optional: Rate limiting req's (reqs per min/hr)
- Optional: Custom middleware / plugin configs
- Optional: TLS certs for HTTPS endpoints
Do
See Extended Examples for complete config files + templates.
Step 1: Install API Gateway
Deploy gateway w/ DB (Kong) or file-based config (Traefik).
For Kong w/ PostgreSQL:
# kong-deployment.yaml (excerpt - see EXAMPLES.md for complete file)
apiVersion: v1
kind: Namespace
metadata:
name: kong
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: kong
namespace: kong
spec:
replicas: 2
# ... (PostgreSQL, migrations, services - see EXAMPLES.md)
For Traefik:
# traefik-deployment.yaml (excerpt - see EXAMPLES.md for complete file)
apiVersion: v1
kind: Namespace
metadata:
name: traefik
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: traefik
namespace: traefik
spec:
replicas: 2
# ... (RBAC, ConfigMap, services - see EXAMPLES.md)
See EXAMPLES.md for complete deployment manifests
Deploy:
kubectl apply -f kong-deployment.yaml # OR traefik-deployment.yaml
kubectl wait --for=condition=ready pod -l app=kong -n kong --timeout=300s
kubectl get svc -n kong kong-proxy # Get load balancer IP
→ Gateway pods running w/ 2 replicas. LB service has external IP. Admin API accessible (Kong: port 8001, Traefik: dashboard port 8080). Health checks pass.
If err:
- Check pod logs:
kubectl logs -n kong -l app=kong - Valid. DB connection (Kong):
kubectl logs -n kong kong-migrations-<hash> - Check service account perms (Traefik):
kubectl get clusterrolebinding traefik -o yaml - Ensure ports not already bound:
kubectl get svc --all-namespaces | grep 8000
Step 2: Configure Backend Services + Routes
Define upstream services + create routes to expose APIs.
For Kong (using decK for declarative config):
# Install decK CLI
curl -sL https://github.com/Kong/deck/releases/download/v1.28.0/deck_1.28.0_linux_amd64.tar.gz | tar -xz
sudo mv deck /usr/local/bin/
# Create kong.yaml with services, routes, upstreams
# (see EXAMPLES.md for complete configuration)
deck sync --kong-addr http://localhost:8001 -s kong.yaml
curl -i http://localhost:8001/routes # Verify routes
For Traefik (using IngressRoute CRD):
# traefik-routes.yaml (excerpt)
apiVersion: traefik.io/v1alpha1
kind: IngressRoute
metadata:
name: user-api-route
spec:
entryPoints: [websecure]
routes:
- match: Host(`api.example.com`) && PathPrefix(`/api/users`)
# ... (see EXAMPLES.md for full configuration)
Apply routes:
kubectl apply -f traefik-routes.yaml
curl -H "Host: api.example.com" https://GATEWAY_IP/api/users
See EXAMPLES.md for complete routing configs
→ Routes proxy traffic correct to backend services. Weighted routing distributes traffic per config. Health checks monitor backend health.
If err:
- Valid. backend services running:
kubectl get svc -n default - Check DNS resolution:
kubectl run test --rm -it --image=busybox -- nslookup user-service.default.svc.cluster.local - Review gateway logs:
kubectl logs -n kong -l app=kong --tail=50 - Valid. config:
deck validate -s kong.yaml
Step 3: Implement Auth
Configure auth plugins/middleware for API security.
For Kong (API Key + JWT auth):
# kong-auth-config.yaml (excerpt)
consumers:
- username: mobile-app
custom_id: app-001
keyauth_credentials:
- consumer: mobile-app
key: mobile-secret-key-123
plugins:
- name: key-auth
service: user-api
# ... (see EXAMPLES.md for full configuration)
deck sync --kong-addr http://localhost:8001 -s kong-auth-config.yaml
curl -i -H "apikey: mobile-secret-key-123" http://GATEWAY_IP/api/users
For Traefik (BasicAuth + ForwardAuth middleware):
# traefik-auth-middleware.yaml (excerpt)
apiVersion: traefik.io/v1alpha1
kind: Middleware
metadata:
name: basic-auth-middleware
spec:
basicAuth:
secret: basic-auth
removeHeader: true
# ... (see EXAMPLES.md for OAuth2, rate limiting)
kubectl apply -f traefik-auth-middleware.yaml
curl -u user1:password https://GATEWAY_IP/api/protected
See EXAMPLES.md for complete auth configs
→ Unauth'd reqs return 401. Valid creds allow access. Rate limiting returns 429 after threshold. JWT tokens valid. correctly. ACL enforces group perms.
If err:
- Valid. consumer creation:
curl http://localhost:8001/consumers - Check plugin enabled:
curl http://localhost:8001/plugins | jq . - Test w/ verbose:
curl -vto see res headers - Valid. JWT: use jwt.io to decode token
Step 4: Configure Request/Response Transformation
Add middleware to transform reqs + res.
For Kong:
# kong-transformations.yaml (excerpt)
plugins:
- name: request-transformer
service: user-api
config:
add:
headers: [X-Gateway-Version:1.0, X-Request-ID:$(uuid)]
remove:
headers: [X-Internal-Token]
- name: correlation-id
# ... (see EXAMPLES.md for full configuration)
deck sync --kong-addr http://localhost:8001 -s kong-transformations.yaml
For Traefik:
# traefik-transformations.yaml (excerpt)
apiVersion: traefik.io/v1alpha1
kind: Middleware
metadata:
name: add-headers
spec:
headers:
customRequestHeaders:
X-Gateway-Version: "1.0"
# ... (see EXAMPLES.md for circuit breaker, retry, chain)
kubectl apply -f traefik-transformations.yaml
curl -v https://GATEWAY_IP/api/users | grep X-Gateway
See EXAMPLES.md for complete transformation configs
→ Req headers added/removed as config'd. Res headers include gateway metadata. Large reqs rejected w/ 413. Circuit breaker trips on repeated fails. Retries occur for transient errs.
If err:
- Valid. middleware order in chain
- Check for header conflicts w/ backend services
- Test transformations individually before chaining
- Review logs for transformation errs
Step 5: Enable Monitoring + Analytics
Configure metrics, logging, dashboards for API visibility.
Kong monitoring setup:
# kong-monitoring.yaml (excerpt)
plugins:
- name: prometheus
config:
per_consumer: true
- name: http-log
service: user-api
# ... (see EXAMPLES.md for Datadog, file-log configuration)
deck sync --kong-addr http://localhost:8001 -s kong-monitoring.yaml
# Deploy ServiceMonitor (see EXAMPLES.md)
kubectl apply -f kong-servicemonitor.yaml
curl http://localhost:8100/metrics
Traefik monitoring (built-in):
# ServiceMonitor (excerpt - see EXAMPLES.md for Grafana dashboard)
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
name: traefik-metrics
spec:
endpoints:
- port: metrics
path: /metrics
interval: 30s
kubectl port-forward -n traefik svc/traefik-dashboard 8080:8080
# Open http://localhost:8080/dashboard/
See EXAMPLES.md for complete monitoring configs
→ Prometheus scraping gateway metrics successfully. Dashboards show req rates, latency percentiles, err rates. Logs forwarding to aggregation system. Metrics segmented by service, route, consumer.
If err:
- Valid. ServiceMonitor:
kubectl get servicemonitor -A - Check Prometheus targets in UI
- Ensure metrics port accessible:
kubectl port-forward -n kong svc/kong-metrics 8100:8100 - Valid. log endpoint reachability
Step 6: Implement API Versioning + Deprecation
Configure version mgmt + graceful API deprecation.
Kong versioning strategy:
# kong-versioning.yaml (excerpt)
services:
- name: user-api-v1
url: http://user-service-v1.default.svc.cluster.local:8080
routes:
- name: user-v1-route
paths: [/api/v1/users]
plugins:
- name: response-transformer
config:
add:
headers:
- X-Deprecation-Notice:"API v1 deprecated on 2024-12-31"
- Sunset:"Wed, 31 Dec 2024 23:59:59 GMT"
# ... (see EXAMPLES.md for v2, default routing, rate limits)
Traefik versioning:
# traefik-versioning.yaml (excerpt)
apiVersion: traefik.io/v1alpha1
kind: Middleware
metadata:
name: v1-deprecation-headers
spec:
headers:
customResponseHeaders:
X-Deprecation-Notice: "API v1 deprecated on 2024-12-31"
# ... (see EXAMPLES.md for complete IngressRoutes)
Test versioning:
curl -i https://api.example.com/api/v1/users # Deprecated
curl -i https://api.example.com/api/v2/users # Current
curl -i https://api.example.com/api/users # Routes to v2
See EXAMPLES.md for complete versioning configs
→ Diff versions route to appropriate backend services. Deprecation headers on v1 res. Rate limits stricter for deprecated versions. Default path routes to latest version. Metrics segmented by API version.
If err:
- Valid. path precedence/priority config (higher priority = eval'd first)
- Check for overlapping path patterns
- Test each version route independently
- Review routing logs for path matching
- Ensure backend services for each version running
Check
- API gateway pods running w/ multi replicas for HA
- LB service has external IP
- Routes proxy traffic correct to backend services
- Auth enforcing access control (401/403 res)
- Rate limiting returns 429 after exceeding quotas
- Req/res transformation adding/removing headers correct
- Circuit breaker trips on repeated backend fails
- Metrics exposed + scraped by Prometheus
- Dashboards showing req rates, latency, errs
- API versioning routing reqs to correct backend versions
- Deprecation headers on older API versions
- Health checks monitoring backend service avail
Traps
-
DB Dependency (Kong): Kong w/ DB requires PostgreSQL/Cassandra. DB-less mode avail but limits features (runtime config changes). Use DB mode for prod w/ multi gateway instances.
-
Path Matching Order: Routes/IngressRoutes evaluated in specific order. More specific paths should have higher priority. Overlapping paths → unpredictable routing. Test w/
curl -vto verify actual route hit. -
Auth Bypass: Ensure auth plugins applied to all routes. Easy to add route w/o auth. Use default plugins at service level, override per-route as needed.
-
Rate Limit Scope: Rate limiting
policy: localcounts per gateway pod. Consistent limits across replicas → use centralized policy (Redis) or sticky sessions. -
CORS Config: API gateway should handle CORS, not individual services. Add CORS plugin/middleware early → avoid browser preflight fails.
-
SSL/TLS Termination: Gateway typically terminates SSL. Ensure certs valid + auto-renewal config'd. Use cert-manager for K8s cert mgmt.
-
Upstream Health Checks: Active health checks detect backend fails quickly. Passive checks rely on real traffic + slower to detect issues.
-
Plugin/Middleware Exec Order: Order matters. Auth before rate limiting (avoid wasted rate limit slots for invalid reqs). Transformation before logging (log transformed values).
-
Resource Limits: Gateway pods can consume big CPU under load. Set appropriate resource reqs/limits. Monitor CPU throttling in prod.
-
Migration Strategy: Don't enable all plugins at once. Roll out incrementally: routing → auth → rate limiting → transformations → advanced features.
→
configure-ingress-networking- Ingress controller setup complements API gatewaysetup-service-mesh- Service mesh provides complementary east-west traffic mgmtmanage-kubernetes-secrets- Cert + credential mgmt for gatewaysetup-prometheus-monitoring- Monitoring integration for gateway metricsenforce-policy-as-code- Policy enforcement complements gateway auth
GitHub リポジトリ
関連スキル
executing-plans
デザインexecuting-plansスキルは、完全な実装計画があり、それを管理されたバッチでレビューチェックポイントを設けながら実行する場合に使用します。このスキルは計画を読み込んで批判的にレビューした後、小さなバッチ(デフォルトは3タスク)でタスクを実行し、各バッチの間に進捗状況を報告してアーキテクトのレビューを受けます。これにより、品質管理チェックポイントが組み込まれた体系的な実装が保証されます。
requesting-code-review
デザインこのスキルは、コードレビュアーサブエージェントを起動し、処理を進める前に要件に対してコード変更を分析します。タスク完了後、主要な機能の実装後、またはmainブランチへのマージ前などに使用すべきです。このレビューは、現在の実装と元の計画を比較することで、問題を早期に発見するのに役立ちます。
connect-mcp-server
デザインこのスキルは、開発者がHTTP、stdio、またはSSEトランスポートを使用してMCPサーバーをClaude Codeに接続するための包括的なガイドを提供します。GitHub、Notion、カスタムAPIなどの外部サービスを統合するためのインストール、設定、認証、セキュリティについて解説しています。MCP統合のセットアップ、外部ツールの設定、またはClaudeのModel Context Protocolを扱う際にご利用ください。
web-cli-teleport
デザインこのスキルは、タスク分析に基づいて開発者がClaude Code WebとCLIインターフェースの選択を支援し、これらの環境間でのシームレスなセッションテレポーテーションを可能にします。Web、CLI、モバイル環境を切り替える際のセッション状態とコンテキストを管理することで、ワークフローを最適化します。様々な段階で異なるツールを必要とする複雑なプロジェクトにご活用ください。
