gke-inference

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill promotes secure practices by instructing users to store sensitive Hugging Face tokens in Kubernetes Secrets rather than hardcoding them within manifests or commands.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for a workflow that incorporates user-provided values into CLI commands, which is a standard pattern for technical manuals but represents a theoretical injection surface.\n
  • Ingestion points: CLI argument placeholders in SKILL.md such as <MODEL_NAME>, <SERVER>, and <ACCELERATOR>.\n
  • Boundary markers: None; parameters are intended for direct interpolation into shell commands.\n
  • Capability inventory: The skill workflow utilizes gcloud and kubectl to manage cluster resources and retrieve logs.\n
  • Sanitization: None explicitly defined within the skill; validation is expected to be handled by the executing agent or the underlying CLI tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 05:41 AM
Security Audit — agent-trust-hub — gke-inference