The Agent Skills Directory

Remote Script Execution: The skill downloads and executes installation and setup scripts from official AWS and OpenTelemetry GitHub repositories.
Evidence: azure-ingestion/custom-telemetry.md uses curl -fsSL https://raw.githubusercontent.com/aws/amazon-cloudwatch-agent/main/scripts/aws/setup.sh | sh to configure web-identity federation.
Evidence: instrumentation/ec2-dotnet.md downloads an installer script from github.com/aws-observability/aws-otel-dotnet-instrumentation/releases/latest/download/aws-otel-dotnet-install.sh and executes it.
Context: These scripts are hosted on vendor-owned or well-known service platforms and are used for their intended purpose of configuring observability agents.
Indirect Prompt Injection Surface: The skill instructs the agent to fetch documentation for various AI frameworks to verify package names and APIs.
Ingestion points: references/cloudwatch-omni/omni-agents-instrumentation/omni-agents-instrumentation.md and related framework guides.
Boundary markers: The skill includes explicit instructions stating that fetched content is untrusted and addressed text must be ignored rather than executed.
Capability inventory: The skill has access to shell commands via aws___call_aws and package managers (pip, npm) for configuring the environment.
Sanitization: Procedural instructions require the agent to report findings to the user rather than acting on instructions found within external documentation.
System Configuration and Privilege Usage: The skill provides commands to install software and manage system services on EC2 instances.
Evidence: Scripts use sudo and systemctl to install the CloudWatch Agent and start collector services.
Context: This usage is necessary for the skill's primary purpose of setting up system-wide telemetry collection on virtual machines.

setting-up-cloudwatch-observability