The Agent Skills Directory

[COMMAND_EXECUTION]: The skill describes the use of shell-based orchestration for tools including the AWS CLI, Terraform, Helm, and kubectl. These operations are core to the skill's primary purpose of provisioning and managing EKS clusters and Keycloak instances.
[INDIRECT_PROMPT_INJECTION]: The design involves ingesting data from external sources such as Kubernetes metrics, Helm release statuses, and tool outputs. While this represents a potential attack surface for indirect prompt injection, the documentation specifically advocates for the implementation of 'Guardrails' and 'Evaluation' loops to validate and verify state before and after execution.
Ingestion points: debug command (tool outputs), Evaluation pattern (metrics endpoints, kubectl logs/status).
Boundary markers: Not explicitly defined in the pattern descriptions.
Capability inventory: Infrastructure provisioning via aws, helm, and kubectl in the setup and ship commands.
Sanitization: The 'Guardrails' enhancement suggests pre-flight checks and RBAC validation, which act as a form of security sanitization for planned operations.

agentic-patterns