ClawHub
Back to skill

Security audit

K8s运维专业版

Security checks across malware telemetry and agentic risk

Overview

This Kubernetes operations skill is mostly purpose-aligned, but it asks for powerful production-cluster command and auto-fix authority with broad triggers and unsupported safety claims.

Review this skill before installing in any environment with real cluster credentials. Use it only with least-privilege kubeconfig contexts, require dry-run and explicit approval for every change, verify any generated command before execution, and avoid enabling scheduled automation or webhooks until the actual implementation and command boundaries are known.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Intent-Code Divergence

Medium
Confidence
84% confidence
Finding
The skill claims command execution occurs in a 'safe sandbox', but the rest of the document only declares generic exec capability and kubectl/system command execution without defining any actual sandbox controls. This can mislead users into granting trust to high-impact command execution that may run with the agent's real privileges against local files and configured clusters.

Vague Triggers

Medium
Confidence
79% confidence
Finding
The natural-language trigger for a full production-cluster health check is broad and maps common operational phrasing directly to powerful actions. In an agent setting, overly permissive invocation language increases the chance of unintended execution against sensitive infrastructure without explicit scoping or confirmation.

Vague Triggers

High
Confidence
95% confidence
Finding
A natural-language request like 'automatically fix all automatically handled issues' is too broad for a destructive or state-changing action. In the context of Kubernetes administration with exec/write tools and multi-cluster support, this could trigger large-scale configuration changes, outages, or unsafe remediation on production resources.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The skill advertises file writing, external service interaction, information retrieval, and command execution as core functions, but the main usage framing does not prominently warn users about those capabilities before invocation. In an agent environment, missing user-facing warnings can cause users to trigger privileged actions without understanding side effects on files, systems, or clusters.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.