Dependencies for the Observability Skill in AWS Agent Toolkit

The observability skill requires no third-party runtime packages or external libraries, depending exclusively on native AWS services (CloudWatch, X-Ray, ADOT, CloudTrail) and optional AWS CDK constructs for infrastructure deployment.

The aws/agent-toolkit-for-aws repository provides this skill as a zero-dependency solution that orchestrates AWS observability services through the standard AWS SDK. Unlike traditional monitoring libraries that require package installation, this skill leverages platform-native capabilities defined in skills/core-skills/aws-observability/SKILL.md, eliminating version conflicts and supply-chain vulnerabilities.

Core AWS Service Dependencies

The observability skill operates entirely within the AWS ecosystem, interfacing directly with managed services rather than external packages. According to the source code in skills/core-skills/aws-observability/SKILL.md, the skill relies on the following platform services:

Amazon CloudWatch

CloudWatch provides the foundational telemetry layer for logs, metrics, alarms, and dashboards. The skill utilizes CloudWatch Logs Insights for query execution, CloudWatch Metrics for numerical data aggregation, and CloudWatch Alarms for threshold-based alerting. This integration requires only standard AWS SDK credentials, not additional software installations.

AWS X-Ray

Distributed tracing capabilities come from AWS X-Ray, which captures request paths across service boundaries. The skill references X-Ray for latency analysis and error tracing, storing segment data in the X-Ray backend without requiring client-side daemons or language-specific agents beyond the AWS SDK.

AWS Distro for OpenTelemetry (ADOT)

The ADOT Collector bridges custom telemetry from applications to CloudWatch. As documented in the routing section of SKILL.md, the collector handles configurable sampling, trace export to X-Ray, and metric export via EMF (Embedded Metric Format). This dependency is satisfied by the managed ADOT service or the optional otel-config.yaml configuration asset.

AWS CloudTrail

For security-oriented observability, the skill integrates with CloudTrail to audit API activity across accounts. This provides governance and compliance visibility without requiring additional logging agents or third-party SIEM connectors.

Optional Infrastructure Dependencies

While the skill itself has zero runtime dependencies, it ships with optional Infrastructure-as-Code assets for rapid deployment. These assets require standard CDK tooling that most AWS projects already maintain.

CDK Alarm Template

The skills/core-skills/aws-observability/assets/alarm-template.ts file provides a TypeScript construct for Lambda monitoring. When using this asset, your project must include the standard AWS CDK dependencies (aws-cdk-lib and constructs), which are typical prerequisites for any CDK-based AWS deployment—not unique to this skill.

// Example: Deploy observability alarms using the CDK helper
import { AlarmTemplate } from './skills/core-skills/aws-observability/assets/alarm-template';

new AlarmTemplate(this, 'LambdaObservability', {
  functionName: 'my-production-function',
  alarmThreshold: 80,
  evaluationPeriods: 2,
});

OpenTelemetry Collector Configuration

The skills/core-skills/aws-observability/assets/otel-config.yaml file provides a sample ADOT collector configuration. This YAML defines receivers for X-Ray traces and exporters for CloudWatch, requiring only the ADOT Collector runtime (managed by AWS) rather than language-specific packages.

# Example: ADOT collector configuration from otel-config.yaml

receivers:
  xray:
    endpoint: 0.0.0.0:2000
    transport: udp
  
exporters:
  awscloudwatch:
    namespace: Application/Observability
    region: us-east-1
    
service:
  pipelines:
    traces:
      receivers: [xray]
      exporters: [awscloudwatch]

Key Implementation Files

Understanding the source structure clarifies where dependencies are defined:

• skills/core-skills/aws-observability/SKILL.md – Master metadata defining service capabilities and routing rules
• skills/core-skills/aws-observability/assets/alarm-template.ts – CDK construct for automated alarm provisioning
• skills/core-skills/aws-observability/assets/otel-config.yaml – Sample collector configuration for custom telemetry pipelines
• skills/core-skills/aws-observability/references/ – Directory containing detailed documentation for each AWS service integration

Summary

• Zero runtime dependencies: The observability skill requires no npm, pip, or Maven packages beyond the standard AWS SDK.
• Native AWS services: Core functionality depends on CloudWatch, X-Ray, ADOT, and CloudTrail platform services.
• Optional CDK assets: Infrastructure templates require standard CDK libraries (aws-cdk-lib, constructs) commonly present in AWS projects.
• Configuration-driven: The otel-config.yaml asset provides telemetry routing without code dependencies.
• SDK-only integration: All service interactions occur through the AWS SDK, which is presumed available in the Agent Toolkit runtime environment.

Frequently Asked Questions

Does the observability skill require installing third-party packages?

No. The skill operates entirely through the AWS SDK and native AWS services. There are no requirements.txt, package.json, or pom.xml entries specific to this skill. The only dependencies are AWS platform services (CloudWatch, X-Ray, ADOT, CloudTrail) that you access via the standard AWS SDK.

Can I use the observability skill without AWS CDK?

Yes. The CDK assets in alarm-template.ts are optional conveniences for infrastructure automation. You can manually configure CloudWatch alarms, X-Ray sampling rules, and ADOT collectors through the AWS Console, CLI, or CloudFormation without installing CDK dependencies.

What AWS permissions are required for the observability skill functionality?

The skill requires standard AWS observability IAM permissions, including cloudwatch:PutMetricData, logs:CreateLogGroup, xray:PutTraceSegments, and cloudtrail:LookupEvents. These align with the AWS managed policies CloudWatchFullAccess and AWSXRayDaemonWriteAccess, though production environments should use least-privilege custom policies.

Is the ADOT Collector mandatory for tracing with this skill?

No. While the otel-config.yaml asset provides ADOT configuration for advanced use cases, the skill also supports direct X-Ray SDK integration. You can send traces to X-Ray using the AWS SDK's built-in tracing capabilities without running the ADOT Collector, making the collector an optional dependency for scenarios requiring custom sampling or multi-backend export.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}