k8s-prometheus-troubleshooting
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill employs
kubectlto perform extensive read-only investigations and state-changing operations likekubectl patchandkubectl label. These modification commands are guarded by "Ask first" instructions to ensure user oversight. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Kubernetes cluster environment and Prometheus API (e.g., target names, error messages, and resource labels) which could contain malicious instructions designed to influence the agent's logic.
- Ingestion points: Cluster data and API responses are retrieved via
kubectl get,kubectl logs, andkubectl get --raw(Prometheus API proxy) as seen inSKILL.mdand the reference documentation. - Boundary markers: The skill contains explicit instructions for the agent to "Read first, ask before changing" and labels specific commands as "Ask first" to prevent autonomous execution of suggestions derived from analyzed data.
- Capability inventory: The agent can execute subprocesses via
kubectlto modify cluster state (SKILL.md). - Sanitization: The skill uses
jqto parse specific fields from API responses, which limits the influence of raw free-text blobs on the agent's context compared to processing unformatted text.
Audit Metadata