k8s-launch-kit-discover

Pass

Audited by Gen Agent Trust Hub on Oct 3, 2026

Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill uses the user's Kubernetes configuration file, typically located at ~/.kube/config, to authenticate with the cluster during the discovery process.
  • [COMMAND_EXECUTION]: The skill executes the l8k discover command-line utility and jq to process and display hardware discovery data.
  • [PRIVILEGE_ESCALATION]: The discovery workflow creates cluster-scoped ClusterRole and ClusterRoleBinding resources named k8s-launch-kit-nic-config-daemon to permit hardware auditing on nodes. These resources are explicitly cleaned up after execution.
  • [EXTERNAL_DOWNLOADS]: The skill bootstraps a DaemonSet which pulls a discovery image from a repository specified in the l8k-config.yaml configuration file.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes hardware metadata from the cluster, such as PCI device IDs and model names, to generate the cluster-config.yaml file.
  • Ingestion points: Hardware attributes are read from sysfs via host pods and NicDevice Custom Resources as described in references/discovery-internals.md.
  • Boundary markers: The skill does not implement explicit delimiters for hardware-derived strings in the generated YAML output.
  • Capability inventory: The skill possesses file system write access and CLI execution capabilities.
  • Sanitization: Hardware identifiers are normalized using lowercase mapping, segment filtering, and deterministic hashing for Kubernetes compatibility.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 3, 2026, 05:22 PM
Security Audit — agent-trust-hub — k8s-launch-kit-discover