Where to Find the Observability Skill File in the AWS Agent Toolkit Repository

The observability skill file is located at skills/core-skills/aws-observability/SKILL.md in the core SDK, with an identical copy packaged as a plugin at plugins/aws-core/skills/aws-observability/SKILL.md.

The AWS Agent Toolkit for aws repository contains declarative skill definitions that power agentic interactions with AWS infrastructure. If you are building agents that monitor, troubleshoot, or optimize cloud resources, locating the observability skill file is essential for understanding its capabilities or extending its functionality.

Core and Plugin Skill File Locations

The toolkit maintains the observability capability in two distinct paths to support different consumption patterns.

Primary Core Skill Definition

The canonical definition resides in the core skills directory at skills/core-skills/aws-observability/SKILL.md. This file serves as the primary source of truth for the observability capability used by the core SDK. It contains the complete YAML metadata definition and trigger keywords that the SkillRegistry parses at runtime.

Plugin-Packaged Version

For agents that load capabilities through the aws-core plugin set, an identical copy exists at plugins/aws-core/skills/aws-observability/SKILL.md. Both files share the same name: aws-observability identifier and contain identical descriptions covering CloudWatch Logs Insights, Metrics, Alarms, Dashboards, EMF, X-Ray, CloudTrail, and ADOT.

Structure of the Observability Skill Definition

The SKILL.md files use YAML frontmatter to declare metadata and routing rules. According to the AWS Agent Toolkit source code, the file defines:

• name: aws-observability
• description: Detailed explanation covering CloudWatch Logs Insights, Metrics, Alarms, Dashboards, EMF, X-Ray, CloudTrail, and ADOT
• triggers: Activation keywords including "cloudwatch", "logs insights", "alarms", "dashboard", "x-ray", and "observability"
• references: Paths to supplemental documentation in the references/ subdirectory

When the registry loads the skill, it indexes these triggers to route natural language requests appropriately.

Loading and Invoking the Skill Programmatically

Agents parse skill files at runtime using the SkillRegistry class. The registry automatically scans repository paths and registers discovered capabilities under their declared names.

from agent_toolkit import Agent, SkillRegistry

# Initialize registry and load core skills including observability

registry = SkillRegistry()
registry.load_from_path("skills/core-skills")

agent = Agent(skill_registry=registry)

# Natural language request triggers the aws-observability skill

response = agent.ask("Help me set up CloudWatch alarms for my Lambda function")
print(response)

In this example, the agent detects the "CloudWatch alarms" trigger phrase and routes the request to the observability skill definition found in skills/core-skills/aws-observability/SKILL.md.

Related Reference Documentation

The observability skill references specialized guides located in the references/ subdirectory relative to the skill file:

• skills/core-skills/aws-observability/references/alarms.md: Configuration details for CloudWatch Alarms
• skills/core-skills/aws-observability/references/troubleshooting.md: Common debugging steps for observability services
• skills/core-skills/aws-observability/references/production.md: Production-readiness checklist

These markdown files extend the core skill definition with implementation-specific guidance that agents consult when handling complex queries.

Summary

• The observability skill file exists in two locations: skills/core-skills/aws-observability/SKILL.md (core SDK) and plugins/aws-core/skills/aws-observability/SKILL.md (plugin)
• Both files contain identical YAML metadata with name: aws-observability and trigger keywords for CloudWatch, X-Ray, and related services
• The SkillRegistry parses these files at runtime to register capabilities and route requests based on trigger keywords
• Supplemental reference materials reside in the references/ subdirectory covering alarms, troubleshooting, and production practices

Frequently Asked Questions

What is the difference between the core and plugin observability skill files?

The content is identical, but the file paths serve different architectural purposes. The skills/core-skills/ version is the canonical definition used by the core SDK, while the plugins/aws-core/ version packages the same capability for agents that consume features through the AWS Core plugin system. Both register under the same aws-observability name within the SkillRegistry.

Which AWS services does the observability skill support?

According to the skill definition in SKILL.md, the capability covers CloudWatch (Logs Insights, Metrics, Alarms, Dashboards, and EMF), X-Ray, CloudTrail, and ADOT (AWS Distro for OpenTelemetry). The triggers include keywords like "cloudwatch", "logs insights", "alarms", "dashboard", "x-ray", and "observability".

How do I extend the observability skill for custom use cases?

You can modify the YAML metadata in skills/core-skills/aws-observability/SKILL.md to add custom triggers or reference additional documentation. You may also create new markdown files in the references/ subdirectory and link them in the references array of the skill definition. Ensure your SkillRegistry loads from the modified path at runtime.

Can agents access the observability skill without loading the entire core skills directory?

Yes. If your agent architecture uses the plugin system, you can load only the aws-core plugin which includes the skill at plugins/aws-core/skills/aws-observability/SKILL.md. Alternatively, instantiate a SkillRegistry and point load_from_path() directly to skills/core-skills/aws-observability/ to load only that specific capability.

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