kubernetes-deploying
Fail
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill includes hardcoded example credentials within a Kubernetes Secret manifest in
SKILL.md. - Evidence of a hardcoded database URL with credentials:
postgresql://user:pass@host:5432/db. - Evidence of a hardcoded API key:
sk-abc123. - [COMMAND_EXECUTION]: The skill provides instructions for executing arbitrary commands within Kubernetes pods and performing administrative cluster operations.
- The
kubectl exec -it <pod-name> -- /bin/shcommand allows for interactive shell access to running containers. - The
kubectl apply -f k8s/command applies configuration changes that can modify cluster security postures. - [INDIRECT_PROMPT_INJECTION]: The skill templates and commands ingest external parameters that are subsequently used in shell execution and manifest application without clear boundaries.
- Ingestion points: User-supplied values for resource names, container images, and configuration data used in the YAML templates and
kubectlcommands inSKILL.md. - Boundary markers: None identified to separate instructions from untrusted data.
- Capability inventory: Arbitrary command execution via
kubectl execand resource modification viakubectl applyinSKILL.md. - Sanitization: No evidence of input validation, escaping, or schema enforcement for user-provided deployment parameters.
Recommendations
- AI detected serious security threats
Audit Metadata