The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several shell scripts (scripts/experiment-runner.sh, scripts/log-collector.sh, scripts/setup-prerequisites.sh) that execute various AWS CLI and kubectl commands.
These commands are necessary for creating experiment templates, starting/stopping experiments, and managing Kubernetes resources.
The setup-prerequisites.sh script specifically creates IAM roles and policies with broad permissions (e.g., ec2:TerminateInstances, rds:FailoverDBCluster, and Resource: "*"). This represents a high-privilege operations surface, although it is documented as a quick-start reference and aligned with the requirements of AWS FIS.
[PROMPT_INJECTION]: The skill possesses a surface for Indirect Prompt Injection (Category 8).
Ingestion points: scripts/log-collector.sh fetches raw application logs from Kubernetes pods and stores them in output/monitoring/step5-logs.jsonl.
Boundary markers: None identified in the provided scripts or instructions.
Capability inventory: The skill has access to powerful tools including aws, kubectl, and arbitrary shell execution through bash.
Sanitization: There is no evidence of log sanitization or filtering before the agent processes this data in Step 6 for analysis.
[EXTERNAL_DOWNLOADS]: The skill references and downloads components from official and well-known sources.
It utilizes MCP servers from the awslabs organization.
It references an optional Chaos Mesh MCP server from a public GitHub repository (RadiumGu/Chaosmesh-MCP) for EKS-based fault injection.

chaos-engineering-on-aws