Back to skill

Security audit

Hologres Privileges

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only Hologres permissions skill with powerful but disclosed admin SQL examples, not hidden or automatic behavior.

Install only if you administer Hologres permissions. Before running generated SQL, verify the target instance, database, schema, role, and user; avoid PUBLIC and SUPERUSER unless explicitly intended; require extra review for ownership transfer, DROP USER/ROLE, REASSIGN OWNED, ALTER DATABASE, and default-privilege changes.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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
Findings (5)

Description-Behavior Mismatch

Medium
Confidence
89% confidence
Finding
The reference goes beyond ordinary grant/revoke guidance and includes ownership transfer plus role creation/deletion, which materially expands the operational scope of the skill. In an agent context, these commands can permanently change administrative control and RBAC structure, creating a larger blast radius than the manifest suggests and increasing the chance of unsafe or over-privileged actions.

Description-Behavior Mismatch

Medium
Confidence
84% confidence
Finding
The reference documents ALTER DATABASE configuration for catalog row-level security, which is a database-wide setting not reflected in the stated privilege-management scope. In an automated skill, hidden database-level changes can affect metadata visibility for all users and complicate troubleshooting or governance if invoked unexpectedly.

Missing User Warnings

Medium
Confidence
86% confidence
Finding
This section documents creation, privilege escalation, superuser assignment, ownership transfer, and user deletion flows without any explicit caution about irreversible impact or least-privilege requirements. In a skill intended to guide operational use, that omission can normalize dangerous administrative actions and increase the chance that users execute destructive or over-privileged commands in production.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill includes examples granting table privileges to PUBLIC and default privileges to PUBLIC without an explicit warning that this effectively exposes data to every database user. In a privilege-management skill, such examples are especially risky because users may copy them verbatim, unintentionally creating broad data exposure across current or future objects.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The section includes destructive or high-impact operations such as ownership transfer, role removal, and role deletion without cautionary guidance. In a privilege-management skill, these actions can revoke effective control, break application access, orphan objects, or unintentionally escalate privileges, especially if an agent proposes or executes them without prominent warnings and confirmations.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.