The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for performing interactive and high-privilege operations within the cluster.
Uses kubectl debug node combined with chroot /host in references/troubleshooting.md to gain direct filesystem access to the underlying Kubernetes host node.
Employs kubectl exec and kubectl debug for interactive shell access to running containers for troubleshooting.
Instructs the use of sudo in references/multi-cluster.md to move downloaded binaries into system-protected directories like /usr/local/bin.
[REMOTE_CODE_EXECUTION]: Several workflows involve downloading and immediately executing scripts from external domains.
references/gitops.md and references/service-mesh.md contain instructions to pipe remote scripts to a shell (e.g., curl | bash) for the installation of Istio, Linkerd, and Submariner.
[EXTERNAL_DOWNLOADS]: The skill fetches tools and configurations from external sources.
Downloads binaries such as clusterctl directly from GitHub releases in references/multi-cluster.md.
Fetches and applies remote Kubernetes manifests from various GitHub organizations including argoproj, bitnami-labs, and kubernetes-sigs.
[CREDENTIALS_UNSAFE]: Example configuration files contain mock sensitive data that may be detected by automated scanners.
references/configuration.md includes an example API key following the sk- pattern (sk-1234567890abcdef) and hardcoded generic passwords in manifest templates.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface when processing external manifests.
Ingestion points: The agent is instructed to process user-provided or external Helm charts, RBAC policies, and YAML manifests in SKILL.md.
Boundary markers: The instructions lack specific delimiters or warnings to ignore embedded instructions within processed data.
Capability inventory: The skill allows for the creation, modification, and deletion of cluster resources, and provides interactive command execution capabilities.
Sanitization: No explicit sanitization or validation steps are provided to check untrusted manifests for malicious instructions before application.

kubernetes-specialist