k8s-helm-troubleshooting

Pass

Audited by Gen Agent Trust Hub on Nov 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill operates by analyzing external data from the Kubernetes cluster environment and user-provided error logs, which constitutes a potential surface for indirect prompt injection if malicious content is present in resource metadata.
  • Ingestion points: Resource names, labels, annotations, and system error strings processed during troubleshooting (SKILL.md, references/install-errors.md).
  • Boundary markers: The skill includes strong instructional safeguards, such as requiring the agent to "Read first, ask before changing" and labeling corrective actions as "Proposed fix" or "Ask first."
  • Capability inventory: Use of kubectl for inspecting cluster state and performing modifications like labeling, annotating, and deleting resources (SKILL.md, references/failed-upgrades.md).
  • Sanitization: The skill relies on human oversight and manual verification of proposed commands as its primary security control.
  • [COMMAND_EXECUTION]: The skill utilizes the kubectl command-line tool to interact with the Kubernetes API. Investigative read operations are used to gather evidence, while mutation commands are explicitly designated for user review and approval before the agent proceeds with execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 8, 2026, 10:58 PM