Nightly Build

Security checks across malware telemetry and agentic risk

Overview

Nightly Build is a coherent maintenance helper, but it asks for unattended scheduled maintenance with update and cleanup authority that is not tightly scoped.

Install only if you are comfortable with a nightly job that can run local commands in your workspace. Keep it report-only or tightly scoped unless you explicitly approve updates and cleanup, inspect the exact commands and paths first, and make sure the cron entry is easy to disable.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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
Findings (2)

Intent-Code Divergence

Medium
Confidence
99% confidence
Finding
This is a true vulnerability in the security reporting logic. The helper `runCmd` catches command failures and returns an `ERROR:` string instead of throwing, so `auditSkills()` will always proceed to mark the skill as secure after `npm audit --audit-level=high`, even when high-severity vulnerabilities are found or the audit itself fails. In a nightly security-reporting skill, this is especially dangerous because it creates false assurance and can systematically hide dependency risk across all scanned skills.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The skill describes automatic updates and cleanup operations that can modify repositories or delete files, but it does not warn users about those side effects or constrain their scope. In an unattended nightly automation context, silent modification/removal of data increases the chance of accidental disruption, loss of logs, or destructive behavior if the task definitions are broadened later.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal