The Agent Skills Directory

[SAFE]: The skill facilitates legitimate Kubernetes cluster management tasks. It incorporates several safety features, such as recommending specific image tags instead of 'latest', implementing liveness and readiness probes, and enforcing resource limits to ensure cluster stability.
[COMMAND_EXECUTION]: The skill correctly identifies and uses established command-line interfaces (kubectl, helm, aws, gcloud) for their intended purposes. It includes standard diagnostic steps like checking rollout status and inspecting pod logs.
[PROMPT_INJECTION]: The skill processes user-provided application metadata (images, replicas, endpoints) to generate manifests. While this represents a data ingestion surface, the instructions emphasize security practices such as using RBAC and dedicated ServiceAccounts which mitigate the risk of unauthorized actions through manipulated inputs.
Ingestion points: User-provided deployment parameters described in the usage section of SKILL.md.
Boundary markers: Not explicitly defined in manifest templates.
Capability inventory: High-privilege cluster management via kubectl, helm, aws, and gcloud tools throughout SKILL.md.
Sanitization: Relies on standard Kubernetes manifest validation and agent-guided code generation.

kubernetes-deployment