The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes Kubernetes manifests and stack configurations that include template interpolation (e.g., {{ .vars.cluster }}), which creates a surface for indirect prompt injection if source files are untrusted.
Ingestion points: The manifests field and file paths defined in the paths attribute of the component configuration.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the provided skill description.
Capability inventory: The skill can perform Kubernetes API operations (apply, delete), Git operations (clone, commit, push), and execute local shell commands via hooks.
Sanitization: The documentation does not specify sanitization or validation of external content before processing.
[COMMAND_EXECUTION]: The skill documentation describes the use of lifecycle hooks (kind: command) which allow for the execution of arbitrary shell commands before or after component operations.
[EXTERNAL_DOWNLOADS]: The skill facilitates the delivery of rendered manifests to external Git repositories (e.g., via the kind: git provision target). References to well-known services like GitHub are used for these GitOps workflows.

atmos-kubernetes