The Agent Skills Directory

[SAFE]: The skill uses official NVIDIA tools (l8k, k8s-launch-kit) and container registries (nvcr.io) to manage networking hardware. The commands and workflows described are standard for Kubernetes infrastructure management.\n- [SAFE]: Instructions mandate the use of --dry-run and l8k validate to verify configurations before application, which helps prevent accidental or malicious misconfigurations.\n- [SAFE]: Data collection via l8k sosreport is used appropriately for cluster diagnostics and troubleshooting, aligning with the senior engineer persona.\n- [SAFE]: The skill encourages structured data processing by instructing the agent to use --output json and parse results with jq, reducing the risk of parsing errors or injection from raw text output.\n- [SAFE]: The skill processes cluster state and diagnostics, which presents a surface for indirect prompt injection. Ingestion points: l8k discover and l8k sosreport outputs (SKILL.md). Boundary markers: None explicitly defined. Capability inventory: l8k deploy (cluster modification) and l8k generate (manifest creation) (SKILL.md). Sanitization: Mandatory JSON output parsing with jq for structure validation.

k8s-network-engineer