k8s-operations
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exposes a significant attack surface by accepting untrusted input for high-privilege operations.
- Ingestion points: The
commandparameter inexec_in_pod, themanifestparameter inkubectl_apply, and thepatchcontent inkubectl_patchare all derived from user-provided strings inSKILL.md. - Boundary markers: There are no explicit instructions or delimiters provided to help the agent distinguish between legitimate administrative commands and malicious instructions embedded within the data it processes.
- Capability inventory: The skill possesses powerful capabilities including
exec_in_pod(running shell commands inside the cluster),kubectl_apply(creating/modifying any resource), anddelete_resource(deleting infrastructure). - Sanitization: The skill lacks instructions for validating or sanitizing user-provided manifests or commands before execution.
- [COMMAND_EXECUTION]: The skill explicitly enables and provides examples for executing shell commands within Kubernetes pods using the
exec_in_podtool. This includes examples usingsh -candcurl, which are powerful primitives that could be abused for cluster reconnaissance or lateral movement if misused.
Audit Metadata