aws-skills_skill

Overview

This skill captures practical AWS development patterns using the CDK, serverless best practices, cost optimization tactics, and event-driven architecture patterns. It focuses on deployable project structure, Lambda and DynamoDB patterns, SQS/EventBridge integrations, security, and monitoring recommendations. Use it as a concise guide when building and operating AWS workloads with JavaScript/TypeScript CDK stacks.

How this skill works

The skill describes a recommended CDK project layout and provides sample stack and construct patterns for API Gateway, Lambda, DynamoDB, S3, SQS, and EventBridge. It explains runtime considerations such as Lambda initialization for connection reuse, single-table DynamoDB access patterns, dead-letter queue handling, lifecycle transitions for S3, and tracing/alarms for observability. It also highlights IAM least-privilege and secrets access patterns to reduce blast radius.

When to use it

• When scaffolding a new CDK project and defining stack/construct boundaries
• When writing Lambda functions that need optimized cold-starts and connection reuse
• When designing DynamoDB single-table access patterns and GSIs
• When implementing asynchronous workflows with SQS or event-driven flows with EventBridge
• When reducing costs with Lambda sizing, Graviton/ARM architecture, and S3 lifecycle rules

Best practices

• Structure code into stacks, reusable constructs, and separate lambda handlers for clarity and reuse
• Initialize clients (DB, caches) outside handlers to enable connection reuse across invocations
• Apply least-privilege IAM grants via resource grant helpers instead of broad policies
• Use dead-letter queues, visibility timeouts, and batching for resilient SQS+Lambda processing
• Prefer pay-per-request for unpredictable traffic, tune Lambda memory/timeout, and use ARM64 for cost savings
• Enable tracing and alarms (X-Ray, CloudWatch) for errors and latency to detect regressions early

Example use cases

• Deploying a REST API backed by Lambda and DynamoDB with CDK-managed stages and throttling
• Implementing a single-table design to store users, orders, and other entities with GSIs for alternate lookups
• Processing order events via EventBridge, invoking multiple targets like Lambda and SQS for fan-out
• Batch-processing messages from SQS with retries and a dead-letter queue for poison message handling
• Applying S3 lifecycle rules to move objects to infrequent access and Glacier to lower storage costs