k8s-autoscaling-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill follows least-privilege principles by focusing on read-only investigation and explicitly warning the agent against accessing sensitive information like Kubernetes Secrets or dumping objects with potentially inline credentials.
  • [INDIRECT_PROMPT_INJECTION]: The skill interacts with external cluster data, creating a surface for indirect prompt injection from log or event content.
  • Ingestion points: Usage of kubectl logs and kubectl get events in SKILL.md and references/metrics-keda-vpa.md to analyze cluster state.
  • Boundary markers: Absent; the skill relies on interpreting specific status reasons and messages within the output.
  • Capability inventory: Cluster modification tools include kubectl patch, kubectl set resources, and kubectl rollout restart as described in SKILL.md and references/metrics-keda-vpa.md.
  • Sanitization: The skill mitigates risks by using specific jsonpath selectors to avoid processing entire object payloads and by explicitly blacklisting Secrets and the use of -o yaml on sensitive objects.
  • [COMMAND_EXECUTION]: The skill generates shell commands for the agent to execute against a Kubernetes cluster.
  • Evidence: Investigation commands are read-only. All modification commands (e.g., kubectl patch) are explicitly categorized as requiring user approval or manual execution, ensuring human-in-the-loop oversight.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 11:22 PM