ClawHub
Back to skill

Security audit

K8s Triage Automator / K8s 故障自动诊断

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only Kubernetes triage skill whose sensitive log handling is disclosed and purpose-aligned, but users should be careful before exporting cluster data to third-party tools.

Before using this with a real cluster, review generated logs and reports for secrets, internal hostnames, tokens, customer data, and stack traces. Use least-privilege Kubernetes access, limit namespaces where possible, and only export to Jira, Slack, PagerDuty, or similar tools after confirming the destination and retention/access controls.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The skill explicitly promotes aggregating logs from failing pods and exporting triage results to Jira, Slack, or PagerDuty, but it provides no warning about sensitive data exposure. In a Kubernetes troubleshooting context, logs and events commonly contain secrets, internal hostnames, tokens, stack traces, and customer data, so encouraging export without data-handling guidance creates a real risk of unintended disclosure.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.