kubectl
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates Kubernetes management by executing the
kubectlCLI tool and several local shell scripts. These scripts automate tasks such as pod debugging (kubectl-pod-debug.sh), deployment updates (kubectl-deploy-update.sh), and node maintenance (kubectl-node-drain.sh). - [EXTERNAL_DOWNLOADS]: The documentation references
clawdhubas a platform for installing and managing this skill package. - [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by reading untrusted content from the cluster environment, such as container logs and system events.
- Ingestion points: The agent reads container logs and cluster events via
kubectl logsandkubectl get events, as seen inSKILL.mdandscripts/kubectl-pod-debug.sh. - Boundary markers: Absent; there are no instructions or delimiters to prevent the agent from obeying potential instructions embedded within the logs of a compromised container.
- Capability inventory: The skill provides broad write and execution capabilities, including
kubectl execfor shell access andkubectl applyfor modifying cluster state, as documented inreferences/REFERENCE.md. - Sanitization: Absent; the skill passes raw log output and event descriptions directly into the agent context without escaping or filtering.
Audit Metadata