kubernetes
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a custom Python wrapper (scripts/k8s-cli) and Bash scripts to execute kubectl and gh commands for cluster operations. These are implemented securely using argument lists to avoid shell injection vulnerabilities.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from cluster resources such as logs and events, which could contain malicious instructions.\n
- Ingestion points: Data is collected from container logs, events, and resource descriptions via scripts/k8s-cli and scripts/gather-cluster-state.sh.\n
- Boundary markers: The k8s-cli tool uses a --json flag to provide structured output, aiding the agent in distinguishing data from control instructions.\n
- Capability inventory: The skill allows full resource lifecycle management through kubectl, including creating and deleting workloads.\n
- Sanitization: The k8s-cli script includes a redaction function that masks common sensitive fields in JSON output to prevent accidental exposure.\n- [DATA_EXFILTRATION]: While the skill accesses sensitive cluster information, it implements specific mitigations such as the redaction of keys like token, password, and secret in the CLI wrapper to prevent data leakage in the agent's context.
Audit Metadata