k8s-api-troubleshooting
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from cluster logs, events, and resource definitions. An attacker with control over these outputs could attempt to inject instructions.\n
- Ingestion points:
kubectl logs,kubectl get events, andkubectl describeinSKILL.mdand reference files.\n - Boundary markers: Instructions explicitly state "Read first, ask before changing" and "Never run a 'Proposed fix'; hand it to the user."\n
- Capability inventory: Access to
kubectl patch,delete, andscalecommands.\n - Sanitization: No explicit sanitization or filtering of log/event content is mentioned before processing.\n- [COMMAND_EXECUTION]: The skill relies on
kubectlcommands to gather information and propose fixes. This is consistent with the skill's purpose and uses standard tool-calling patterns.
Audit Metadata