k8s
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill accesses the Kubernetes configuration file (~/.kube/config) and the KUBECONFIG environment variable, which store sensitive authentication tokens and cluster certificates for the targeted environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes information fetched from the Kubernetes API, which creates an attack surface where malicious cluster content could influence the agent.\n
- Ingestion points: Fetches node details, pod summaries, and event logs in scripts/discover.ts.\n
- Boundary markers: Absent; no instructions are provided to help the agent distinguish between legitimate command output and potentially malicious text embedded in resource names or event messages.\n
- Capability inventory: The agent is authorized to execute a range of kubectl and helm commands as specified in SKILL.md.\n
- Sanitization: Absent; while jq is used for structural extraction, no sanitization is performed on natural language fields like event messages.\n- [COMMAND_EXECUTION]: Instructions in SKILL.md direct the agent to execute shell commands using background jobs and parallel processing patterns.\n- [DYNAMIC_EXECUTION]: The discovery script (scripts/discover.ts) utilizes Bun's shell implementation to execute kubectl binaries with dynamically constructed arguments.
Audit Metadata