AWS Agentic AI Plugin

The aws-skills repository provides specialized knowledge and MCP (Model Context Protocol) server integrations for Claude Code.

Available Plugins

• AWS Common Plugin: A required dependency providing AWS Documentation MCP configuration for up-to-date knowledge.
• AWS CDK Plugin: Focuses on Infrastructure as Code (IaC) with integrated CDK MCP servers and best practices.
• AWS Cost & Operations Plugin: Provides cost estimation, optimization, and monitoring via integrated MCP servers for AWS Pricing, Cost Explorer, and CloudWatch.
• AWS Serverless & Event-Driven Architecture Plugin: Covers Well-Architected serverless design, Step Functions orchestration, and patterns like Saga or Event Sourcing.
• AWS Agentic AI Plugin: An expert for Amazon Bedrock AgentCore, covering gateway services, runtime, memory, identity, and observability for AI agents.

The aws-cost-operations skill is designed for AWS cost optimization, monitoring, observability, and operational excellence. It is the primary skill to use when you need to analyze AWS bills, estimate costs for new resources, set up CloudWatch alarms, query logs, audit CloudTrail activity, or assess security posture.

Key triggers for this skill include mentions of:

• AWS costs, spending, billing, budget, or pricing
• CloudWatch, observability, monitoring, or alerting
• CloudTrail, audit, or security posture
• Optimizing infrastructure costs or operational efficiency.

SST v4 (the "Ion" engine) is a Pulumi-backed Infrastructure as Code (IaC) framework where you describe AWS resources using TypeScript. It provides high-level sst.aws.* components (such as Function, Bucket, Dynamo, Cron, Service, etc.) that expand into multiple underlying AWS resources. It also allows for an escape hatch to use any raw Pulumi aws.* resources.

Key Identification:

• Use this skill when working with sst.config.ts, $config, $transform, $interpolate, sst.aws.*, sst.Secret, or running commands like sst deploy, sst dev, sst diff, or sst remove.
• Note: This is specifically for SST v4. For SST v2/v3 (SST Classic/CDK-based), these patterns do not apply.

AWS Bedrock AgentCore is a platform for deploying and scaling AI agents using nine core services. Use the following table to identify the service required for your task:

For specific implementation details, refer to the individual service READMEs within the services/ directory.

Pattern B is a stateless approach where the Agent uses local @tool functions to make direct HTTP calls to backend APIs. This pattern does not use the MCP protocol and has no persistent connection overhead. It is ideal for calling internal microservices (e.g., /api/svc/*).

Each tool call is an independent request. You can implement wrappers to automatically inject authentication headers (like API keys) into these requests.

@tool
def http_request(url: str, method: str = "GET", **kwargs) -> str:
    """Call backend API with auto-injected auth headers."""
    headers = kwargs.get("headers", {})
    if "/api/svc/" in url:
        headers["X-Service-Api-Key"] = _service_api_key
    response = requests.request(method, url, headers=headers, **kwargs)
    return response.text

Avoid 'Lambda chaining' (where one function invokes another directly). Instead, use AWS Step Functions to orchestrate workflows. This provides visual representation, built-in error handling/retries, execution history, and service integrations without writing glue code.

// ❌ BAD - Lambda function chaining
export const handler1 = async (event: any) => {
  const result = await processStep1(event);
  await lambda.invoke({
    FunctionName: 'handler2',
    Payload: JSON.stringify(result),
  });
};

// ✅ GOOD - Step Functions orchestration
const stateMachine = new stepfunctions.StateMachine(this, 'OrderWorkflow', {
  definition: stepfunctions.Chain
    .start(validateOrder)
    .next(processPayment)
    .next(shipOrder)
    .next(sendConfirmation),
});