model-serving-kubernetes
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
kubectlandhelmcommands to manage Kubernetes resources, install the KServe controller, and deploy model serving manifests. - [EXTERNAL_DOWNLOADS]: Fetches deployment artifacts from well-known sources, including the KServe Helm repository and official container images from the vLLM project and NVIDIA Container Registry.
- [PROMPT_INJECTION]: An indirect prompt injection surface is identified.
- Ingestion points: Model inference endpoints defined in
SKILL.md(e.g., the iris and llama-3-8b services) ingest untrusted data from external sources for processing. - Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are defined for the inference data payloads.
- Capability inventory: The skill possesses the capability to modify cluster state and run commands via
kubectlandhelmas described inSKILL.md. - Sanitization: No input validation or sanitization logic is present in the provided Kubernetes manifests or example commands.
Audit Metadata