k8s-workload-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 processes untrusted data from Kubernetes logs and events, which serves as a potential vector for indirect prompt injection.
  • Ingestion points: The skill ingests untrusted data via kubectl logs, kubectl events, and kubectl describe commands (referenced in SKILL.md, references/crashloop.md, references/probes.md, and references/rollouts.md).
  • Boundary markers: There are no explicit delimiters or boundary markers used when interpolating log or event content into the agent's context.
  • Capability inventory: The skill utilizes powerful modification capabilities including kubectl set image, kubectl set resources, kubectl patch, kubectl rollout undo, and kubectl exec (found in SKILL.md and reference files).
  • Sanitization: The skill provides manual sanitization instructions, telling the agent to "Never read Secrets" and "If logs or output contain passwords, tokens or keys, do not repeat them".
  • [COMMAND_EXECUTION]: The skill relies on executing kubectl commands for investigation and remediation.
  • The skill uses kubectl get, describe, logs, events, and rollout for investigation.
  • Modification commands like kubectl set image, patch, and rollout undo are explicitly gated behind user approval ("Ask first" pattern), reducing the risk of unauthorized changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 8, 2026, 10:57 PM