gke-manifest-generation

Pass

Audited by Gen Agent Trust Hub on Oct 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • Command Execution via CLI Tooling: The skill instructs the agent to use the gcloud CLI tool to generate manifests for AI workloads. This involves interpolating user-provided parameters (such as {model_name}) into a shell command string. While this is an intended feature for manifest generation using vendor-official tools, it presents a potential command injection surface if the input parameters are not appropriately sanitized by the agent platform.
  • Indirect Prompt Injection Surface: The skill is designed to ingest natural language descriptions or application code changes to produce Kubernetes manifests. This provides an entry point for untrusted data that could attempt to influence the agent's output. The risk is minimized as the skill primarily outputs static configuration files and provides strong hardening rules (e.g., non-root execution, read-only filesystems) to enforce a secure baseline.
  • Dynamic Execution Pattern: The generation of manifests via the gcloud container ai profiles manifests create command represents a dynamic execution pattern where resource configurations are computed at runtime based on specific user requirements. This is a common pattern for cloud infrastructure tooling but warrants review to ensure inputs are handled securely.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 19, 2026, 07:05 PM