rancher-resource-troubleshooting

This skill is a plausible, legitimate Rancher/Kubernetes troubleshooting skill: its stated capabilities (logs, events, describe, get, and delegating deeper analysis to specialized sub-agents) match common troubleshooting workflows. It does not contain explicit malicious code, external downloads, or obfuscated payloads. However, it enables powerful read access to cluster resources (logs and manifests) and delegates work to sub-agents, creating a transitive trust and data-exposure risk. The primary concerns are: (1) sub-agent transitive access (skill-to-skill trust), (2) potential exposure of secrets or sensitive data via logs/manifests, and (3) broad/parallel operations without explicit per-action consent or redaction rules. I assess low probability of intentional malware (malicious payloads or exfiltration instructions are not present) but a moderate security risk due to capability scope and delegation. Recommend ensuring least-privilege MCP credentials, auditing and restricting sub-agents, redaction of sensitive fields in returned manifests/logs, and user confirmations for wide-scoped actions.