The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts (scripts/cleanup_idle_resources.py) that use the boto3 library to interact with AWS APIs. These scripts are capable of deleting unattached EBS volumes and old snapshots. This functionality is explicitly documented as the primary purpose of the script for cost optimization and includes a --dry-run safety mechanism.
[EXTERNAL_DOWNLOADS]: The GitHub Actions workflow (examples/ci-cd/infracost-github-action.yml) and Terraform configurations reference official actions and providers from well-known sources (HashiCorp, Infracost, and GitHub). These are standard dependencies for infrastructure management and cost estimation.
[CREDENTIALS_UNSAFE]: The skill uses placeholders for sensitive information, such as YOUR_TOKEN_HERE in examples/kubernetes/kubecost-values.yaml and CHANGE_ME_PLEASE for administrative passwords. It correctly follows best practices by using secrets management (e.g., secrets.INFRACOST_API_KEY) for CI/CD pipelines.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface where the agent processes cloud infrastructure metadata (volume status, tags, utilization metrics). While this data could theoretically be manipulated to influence agent decisions, the provided scripts perform direct programmatic operations with no natural language interpretation of the external data, significantly reducing the risk.

optimizing-costs